| ||||||
Description: CISA's Emergency Directive to ALL Federal agencies regarding SharePoint. NVIDIA firmly says "no" to any embedded chip gimmicks. Dashlane is terminating its (totally unusable) free tier. Malicious repository libraries are becoming even more hostile. The best web filter (uBlock Origin) comes to Safari. The very popular SonicWall firewall is being compromised. More than 100 models of Dell Latitude and Precision laptops are in danger. The significant challenge of patching SharePoint (for example). A quick look at my DNS Benchmark progress. Does InControl prevent an important update? An venerable sci-fi franchise may be getting a great new series. What to do about the problem of AI "website sucking"?
High quality (64 kbps) mp3 audio file URL: http://media.GRC.com/sn/SN-1038.mp3 |
SHOW TEASE: It's time for Security Now!. Steve Gibson's here. The government has something to say to federal agencies regarding SharePoint. Oh, and there's another, oh my god, Exchange flaw you need to know about. We're going to talk about the best adblocker finally coming to Safari, a very popular firewall that has been compromised, and why, if you have a Dell Latitude or Precision laptop, you must listen to this show. All that and more coming up next on Security Now!.
| Leo Laporte: This is Security Now! with Steve Gibson, Episode 1038, recorded Tuesday, August 12th, 2025: Perplexity's Duplicity. It's time for Security Now!. I know you wait all week for this. It's Tuesday, woohoo. The best part of Tuesday is when Steve Gibson shows up to fill us in on what horrible things have been happening all week. But, you know, Steve, it's funny because it doesn't scare me. It reassures me there's some sanity in all this, that there's somebody thinking about this. And even if we're decrying Microsoft's, you know, inability to write safe software or, you know, this plague of ransomware, it feels like it's being managed by this guy right here. So thank goodness for Steve. |
| Steve Gibson: Well, and actually not by me. I'm just reporting on it. But I would argue that all the problems which have been created have created another industry of their own, all these security researchers who are digging into this stuff. We'll be talking today about Cisco's Talos Group, who went, like, way beyond the call of duty, reverse engineering a chip in Dell laptops. And, you know, thanklessly. They didn't know what they were going to find. What they found was horrifying. |
| Leo: Oh. |
| Steve: As it turns out. And it's, like, Dell, you know, made it more difficult. Somewhere Dell has a file of all the source code of the firmware in this security chip. But do they give it out? No, it's proprietary. But thankfully the laws have indicated that reverse engineering something you own is within your legal rights. And so here they, like, find this massive problem in more than 10 million, more than 100 models of Dell laptops and say, oh, by the way, Dell, here you go. |
| Leo: Here's the fix. |
| Steve: Yeah. Like, thanklessly, essentially. |
| Leo: Yeah, that's pretty amazing. |
| Steve: So there is this huge industry that has basically sprung up to deal with some of the consequences of the fact that, boy, we just jumped into all this without really knowing how to write code that was bug free. You know, once upon a time when it was your word processor, and the cursor would disappear... |
| Leo: Didn't matter. |
| Steve: ...it'd be like, oh, well. And you'd shut it down and start up again and get your cursor back. Now, you know, something crawled into your network and erased all your data. So it's a different scale of problems. |
| Leo: What a world. |
| Steve: So it's a different scale of problems. Anyway, we're going to have fun today because today's topic for Security Now! Episode 1038 - and you're going to check Leo, but I had it in my brain last time we talked about this that between this Tuesday and next Tuesday is the anniversary of this podcast. |
| Leo: The first Security Now! was August 18th, 2005. |
| Steve: Yes. So it will be before we're here again on the 19th. |
| Leo: Wow. |
| Steve: Will be the anniversary. So that means this is the last podcast of our 20th year, and we'll be starting into Year 21, finally all grown up. |
| Leo: We can drink. |
| Steve: That's right. We don't have to, when we press the "I'm 21" button now, it's the truth. |
| Leo: Wow. By the way, you can, if you go to our website, we've got all 1038 episodes, including this, the very first one from August 18th. It's kind of fun to listen to the first one, I have to say. We've come a long way since then, let's put it that way. |
| Steve: Oh, well, yes. And I'm - yeah. We've figured out how to do this. So today's topic is Perplexity's Duplicity. |
| Leo: Oh, I really wanted you to talk about this. Okay. |
| Steve: Oh, am I going to. |
| Leo: Because it was a complicated story, and I wasn't sure who to believe, to be honest. |
| Steve: It was a complicated story, but it's actually, it's part of a much bigger story. And so I'm going to spend some time before we get into the details of what Cloudflare found, setting a larger context because, as I did this, this just started out as a news blurb at the top of the show. And as I dug into it more, it became our topic for today. But first we're going to talk about CISA's emergency directive to all federal agencies regarding SharePoint. NVIDIA firmly saying no to the notion of any embedded chip gimmicks. Dashlane is terminating its free tier. And when I looked at this, I couldn't believe how limited the free one was. First I was thinking, oh, I'm going to, you know, make sure our listeners know. There's no way anybody could actually be using it. So it's kind of a non... |
| Leo: Big deal, in other words. |
| Steve: Yeah, exactly. Also we've got malicious repository libraries becoming even more hostile. The best web filter has come to Safari. A very popular, or, well, the very popular SonicWall Firewall is being compromised, and that's not good. Also, as I mentioned, it turns out that more than 100 different models of Dell's Latitude and Precision laptops are in serious danger, so definitely any of our listeners who are Dell Latitude laptop users need to go to the link that I've got in the show notes. Oh, and one of our listeners is an enterprise IT guy who I mentioned last week because he wrote me an updated email saying, hey, I hear that you're going to talk about me this week. Here's a little more. And what he wrote was so much more than I had to excerpt from it. But a really good look into the behind-the-scenes, what it actually takes to keep Microsoft stuff synchronized. And it turns out it's not as easy as just saying, oh, let's apply a patch to SharePoint. The ripple effects are astonishing. Also we're going to, as a consequence of another listener question, take a look at where I am with the DNS Benchmark because I want to share some of the things that have been happening over the last six months. Turns out that maybe InControl, my InControl freeware may be preventing some updates which people may want, or may not. Also tonight a venerable sci-fi franchise is getting a new series that looks quite hopeful. We're going to talk about that, and then take a look at the generic problem of AI website sucking in the context of and basically put on the map because of what Cloudflare has documented about what they discovered that Perplexity was doing. |
| Leo: Oh, interesting. |
| Steve: So a really, really, really good podcast. I think maybe, Leo, we've got the hang of it now. |
| Leo: After 1038 episodes in 20 years, we've finally learned. |
| Steve: How to do this. |
| Leo: How to do a podcast. |
| Steve: Our great friend Alex Neihaus used to... |
| Leo: Yeah, our first advertiser, yes. |
| Steve: And he used to send me mail saying, Steve, it's better than ever. I think he finally got tired of saying that, or felt that maybe it was redundant. |
| Leo: He still listens, I know, because we hear from him all the time, yeah. Hi, Alex. Good. All right. So a big show coming up, and of course our Picture of the Week, which I have, as always, sealed myself in a soundproof booth, and I have not looked. |
| Steve: I had so much fun with this one. Yeah. |
| Leo: Well, I'm going to get my... |
| Steve: Tell us why we're here, and then we'll do that. |
| Leo: Yes. Well, we're here because of you, Steve. But because we don't want Steve to work for free, we also have some sponsors and the Club. And we have... |
| Steve: Hear the price of coffee's going up. |
| Leo: A lot. A lot. I'm not happy about that. Now is the moment on Security Now! I always look forward to. |
| Steve: Okay, now, I gave this picture the caption "The latest solution for controlling the high cost of healthcare." |
| Leo: Okay. |
| Steve: The latest solution for controlling the high cost of healthcare. |
| Leo: And we are going to now scroll up and see what that solution is, as soon as I get my mouse to scroll. Scroll, mouse. You know what, I laugh, but this is exactly the sign that I see every time I go to the doctor. You want to read it for us, Steve? |
| Steve: Yeah, this is a big sign. Again, the latest solution for controlling the high cost of healthcare. This is a large sign that says: "Stop. Please do not enter our hospital if you're feeling unwell." |
| Leo: Only healthy people in the hospital, please. |
| Steve: That's right. If you're feeling great, you don't have any problems or complaints, come right on in. |
| Leo: It does cut costs considerably in the medical profession. |
| Steve: I just love that. |
| Leo: That's so great. |
| Steve: And I gave it the little subhead "Sometimes the best solutions are the most obvious." |
| Leo: Yes. |
| Steve: So, you know. Why didn't I think of that? Right. Okay, so last Thursday the 7th, CISA issued a rare emergency directive ordering all - and in the reporting "ALL" was in all caps - all federal agencies to patch a new attack vector in Microsoft Exchange email servers - so this is in Exchange, not in SharePoint, Microsoft Exchange. And, okay, so this was Thursday the 7th, gave them just four days, which included the two days over the weekend, in which to do so. They had until Monday morning to get this done. Okay. So let's step back for a moment and examine CISA's authority to compel the actions of federal agencies. It turns out the authority does exist, and it comes from section 3553(h) of title 44 of the U.S. Code, which says: "Authorizes the Secretary of Homeland Security, in response to a known or reasonably suspected information security threat, vulnerability, or incident that represents a substantial threat to the information security of an agency, to 'issue an emergency directive to the head of an agency to take any lawful action with respect to the operation of the information system, including such systems used or operated by another entity on behalf of an agency, that collects, processes, stores, transmits, disseminates, or otherwise maintains agency information, for the purpose of protecting the information system from, or mitigating, an information security threat.'" So indeed, this is in the law. CISA is able, acting under the auspices of the Secretary of Homeland Security, to tell all federal agencies they have to do something. So CISA's emergency directive in this case explains what's going on. It says: "CISA is aware of a post-authentication vulnerability in Microsoft Exchange hybrid-joined configurations that allows an attacker to move laterally from on-premises Exchange to the M365 cloud environment." They said: "This vulnerability poses grave risk to all organizations operating Microsoft Exchange hybrid-joined configurations that have not yet followed the April 2025 patch guidance, and immediate mitigation is critical." They said: "Although exploitation of this vulnerability is only possible after an attacker establishes administrative access on the on-premises Exchange server, CISA is deeply concerned at the ease with which a threat actor could escalate privileges and gain significant control of a victim's M365 Exchange Online environment." So as I mentioned, the emergency directive issued on Thursday begins with "By 9:00 AM EDT on Monday, August 11th, 2025, ALL" - that's where the caps are - "agencies must." And then it goes on. So that was yesterday, you know, based on when we're recording this podcast. And then it proceeds to enumerate a rather long list of steps that must be taken and reported back to CISA upon their completion. This all surrounds an Exchange Server, turns out not really a bug, a design flaw which affects this hybrid on-prem/cloud environment where Exchange on-premise servers sync their data to an Exchange Online, you know, cloud-based instance. Microsoft explained that in old default setups, prior to this April 2025 update to the architecture, on-prem servers shared the authentication service, known as Service Principal, with the synced online instance. When deployed for the first time, default hybrid installations would upload the on-prem authentication certificates to the Service Principal to allow local instances of Exchange to authenticate with the Exchange Online server in order to sync their data. The problem arises when attackers compromise an on-premises Exchange server because they can abuse their control over the system to hijack, create, or alter authentication tokens that grant intruders access further up the cloud environment. And what's more, Microsoft says this attack scenario does not leave an "easily detectable and auditable trace," meaning that this could all be invisible to compromised companies. Look at your logs, you know, there's just no indication this has been going on. This is not something Microsoft had thought to log. So owners of on-prem servers must install the April hotfix - and as I said, a little bit later we'll take a look at why that's not as easy as it may sound - which converts the connection between on-prem and online environments. Instead of being a direct connection, it converts it into a standalone Entra app. It's then necessary to follow setup instructions that include steps to clean those older hybrid authentication certificates and upload new ones to the separate Entra hybrid app exclusively, where an attacker's access is then much more limited, you know, limited the way Microsoft would like it to be. Because this hybrid on-prem-to-online attack, essentially a bridge, is essentially, as I said before, a design flaw, it not only works on Exchange 2016 and Exchange 2019, but also on the latest pay-as-you-go Exchange Server Subscription Edition. So customers of all three versions will have to follow the steps if they are using any sort of a hybrid configuration. So I just wanted to make sure that all of our listeners knew that. |
| Leo: To be clear, this is a brand new Exchange vulnerability. |
| Steve: Yes. |
| Leo: Any Microsoft software you're not updating assiduously, you're in deep trouble. Even if you are, you might be. |
| Steve: Well, and boy, by the time we're done here, Rosco is this listener who does this. And he said, let me give you, like, let me explain why people are not updating, you know. |
| Leo: Ah. |
| Steve: Basically he was a little defensive. It's like, you know, you say we ought to all update SharePoint. Let me tell you why we're not updating SharePoint. |
| Leo: Right, right. |
| Steve: And so it turns out there's a big story. Anyway, what this means for any of our listeners who may not be within federal agencies, but may still be responsible for their enterprise's hybrid on-premises/cloud Exchange Server setup, is that all of this must be done to be secure. Anyway, for anyone needing more information, I have links to CISA's directive and Microsoft's disclosure in the show notes, two links there. It's something that needs to get done. Again, it does require that you basically replace the previous architecture with an Entra-based architecture, which fixes the previous way that on-prem authenticated itself to Exchange Server in the cloud. The problem was that Exchange Server in the cloud would receive the on-prem authentication certificates which the bad guys may have gotten a hold of, and then they would have free rein throughout your whole cloud environment. |
| Leo: Microsoft does not want you to run stuff on-prem, do they. |
| Steve: Actually, no. They are... |
| Leo: That's really the truth. |
| Steve: I think, well, and you know, there was something we touched on in passing, or maybe it's something that I saw doing my research for the podcast a couple weeks ago, but their stock price, Microsoft's stock price is specifically doing as well as it is because of the clear future revenue they are predicting from moving every - moving their world to a subscription model. |
| Leo: But we knew this for a decade that this was the future, the model. |
| Steve: Yes, the plan. And now here it is. Welcome to the future. Here we are today. NVIDIA firmly said "no." Various news outlets have reported that U.S. officials have been exploring the idea of mandating - well, of course, now we know that NVIDIA decided to buy their access to China for 50% of their proceeds, but that's another matter. They've been exploring the idea of mandating that NVIDIA include kill switches, backdoors, and location-tracking tech in its chips to prevent products from reaching adversaries like China. Okay? This pushing, and the rumors that it inevitably creates, you know, would clearly have a negative impact on Nvidia's sales. I mean, already China is concerned about, you know, chips coming from major U.S. suppliers, in the same way that we're concerned about chips coming from China. I mean, that's reciprocal. |
| Leo: Right, yeah. |
| Steve: So this led NVIDIA to make the following formal statement for the record. Last week they posted under the title "No Backdoors. No Kill Switches. No Spyware." They said: "NVIDIA GPUs are at the heart of modern computing. They're used across industries from healthcare and finance to scientific research, autonomous systems, and AI infrastructure. NVIDIA GPUs are embedded into CT scanners and MRI machines, DNA sequencers, air-traffic radar tracking systems, city traffic-management systems, self-driving cars, supercomputers, TV broadcasting systems, casino machines, and game consoles." |
| Leo: Everything. |
| Steve: That's right. Probably in a machine that's some widget you have near you. And Leo, you're probably wearing four of them. |
| Leo: I'm wearing three of them around my neck right now. |
| Steve: That's right. |
| Leo: Yes. |
| Steve: They said: "To mitigate the risk of misuse, some pundits and policymakers propose requiring hardware kill switches or built-in controls that can remotely disable GPUs without user knowledge and consent. Some suspect they might already exist. NVIDIA GPUs do not and should not have kill switches and backdoors. Hard-coded, single-point controls are always a bad idea. "NVIDIA has been designing processors for over 30 years. Embedding backdoors and kill switches into chips would be a gift to hackers and hostile actors. It would undermine global digital infrastructure and fracture trust in U.S. technology. Established law wisely requires companies to fix vulnerabilities, not to create them. Until recently, that policy was universally upheld and beyond question. When security researchers discovered vulnerabilities such as 'Spectre' and 'Meltdown' for CPUs, governments and industry responded with speed and unity to eliminate the risk. "That principle still holds. There is no such thing as a 'good' secret backdoor only dangerous vulnerabilities that need to be eliminated. Product security must always be done the right way: through rigorous internal testing, independent validation, and full compliance with global cybersecurity standards. Robust security is built on the principle of 'defense in depth': layering multiple safeguards so that no single point of vulnerability can compromise or shut down a system. For decades, that's how NVIDIA and American industry have promoted innovation while protecting users and growing the economy. This is no time to depart from that winning formula. "The Clipper Chip was a debacle, a policy and technical failure. The cybersecurity community learned these lessons the hard way during the 1990s with the NSA's Clipper Chip initiative. Introduced in 1993, the Clipper Chip was designed to provide strong encryption while maintaining government backdoor access through a key escrow system. "The Clipper Chip represented everything that's wrong with built-in backdoors. Security researchers discovered fundamental flaws in the system that could allow malicious parties to tamper with the software. It created centralized vulnerabilities that could be exploited by adversaries. The mere existence of government backdoors undermined user confidence in the security of systems. "Kill switches and built-in backdoors create single points of failure and violate the fundamental principles of cybersecurity. We must promote smart software tools, not dangerous hardware traps. Some point to smartphone features like 'find my phone' or 'remote wipe' as models for a GPU kill switch. That comparison doesn't hold water. Optional software features, controlled by the user, are not hardware backdoors. "NVIDIA has always supported open, transparent software that helps customers get the most from their GPU-powered systems diagnostics, performance monitoring, bug reporting, and timely patching with the user's knowledge and consent. That's responsible, secure computing. It helps our customers excel and our industry to stay ahead. "Hardwiring a kill switch" - in case anybody at this point had any doubts. "Hardwiring a kill switch into a chip is something entirely different: a permanent flaw beyond user control, and an open invitation for disaster. It's like buying a car where the dealership keeps a remote control for the parking brake, just in case they decide you shouldn't be driving." |
| Leo: Actually, most cars you buy today have that. I hate to say it. |
| Steve: Bad example, maybe. |
| Leo: Yeah. |
| Steve: He says: "That's not sound policy. It's an overreaction that would irreparably harm America's economic and national security interests. Hardware integrity must be nonpartisan and nonnegotiable. For decades, policymakers have championed industry's efforts to create secure, trustworthy hardware. Governments have many tools to protect nations, consumers, and the economy. Deliberately weakening critical infrastructure should never be one of them. There are no back doors in NVIDIA chips. No kill switches. No spyware. That's not how trustworthy systems are built, and never will be." |
| Leo: Right on. They're right. |
| Steve: Yeah. |
| Leo: Do you believe them? |
| Steve: Yeah, oh, I do, yeah, because we know him. |
| Leo: Yeah, that's Jensen, yeah. |
| Steve: Yes. He's the real deal. And what I like is that this really - I hope that politicians are aware of this and listen to this because of course his bringing up the Clipper Chip was brilliant because it was a complete fiasco. |
| Leo: Yeah, what a flop that was, yeah. |
| Steve: Yes. And this is what the UK and the EU are still trying to do. |
| Leo: Yeah. |
| Steve: We know, like, this battle hasn't yet been resolved officially. And I'm just glad that NVIDIA has planted a very clear, hopefully, stake through the heart of this thing because this whole notion of the government needs its own special access, that just has to be put down. |
| Leo: It's a terrible idea, yeah. |
| Steve: Yeah. |
| Leo: Agree 100%. |
| Steve: Unfortunately, at this point we're still seeing no apparent end in sight. Something's going to happen in October in the EU, so everyone's waiting to see whether this, you know, the pre-encryption filtering of phone content is what the EU tries to do. I just think everybody, all the players, all the technology players, have to, as a unit, say no. |
| Leo: No. |
| Steve: And it's like, if you'd like to have no technology in your union, fine. We're not going to play. I was going to mention, as I mentioned at the top of the show to our listeners, that the Dashlane password manager would be ending its free-tier service coming up on September 16th, so next month, middle of next month. And I figured I'd aim anyone who might be using Dashlane, inertia being what it is, over at either of this network's two password manager sponsors, 1Password or Bitwarden. But then I saw Dashlane boasting that their paid Premium plan, which will be the only thing left - well, they have a Family plan, too - would allow for [get this, Leo, it's an amazing feature] unlimited passwords and passkeys. |
| Leo: Wow. How do they do that? |
| Steve: I know. It's unlimited. And that of course caused me to wonder if unlimited passwords and passkeys is the big selling point of Dashlane's Premium plan, what could possibly be the limit they had imposed for their free plan? |
| Leo: Yeah? |
| Steve: Believe it or not, I'm not kidding you, you get all of 25 passwords for free with Dashlane. |
| Leo: Twenty-five. |
| Steve: Well, you used to up until this coming September. 25! In the year of our lord 2025 it's not possible to meaningfully use any password manager that imposes any limit - well, okay, unless it's 1,000 - on the number of passwords it will store. But, you know, under their "Why upgrade to Premium?" explanation, Dashlane actually says this. They say: "Unlimited passwords and passkeys: Say goodbye to the 25-password limit and start saving every password." |
| Leo: Every one? Wow. |
| Steve: Yes. Anyone you could want, Leo, you could save. |
| Leo: Wow. |
| Steve: But only if you upgrade, not for free. You know, and when you think about it, Leo, how convenient would it then be not to need to decide which of the precious 25 passwords you'll choose to save in your Dashlane vault. |
| Leo: Well, there's an easy fix. Just use the same password everywhere, and then you don't have to... |
| Steve: You don't need - well, that's true. But you don't need to delete any of those less needed passwords to make room for that 25th one that now it's more important. |
| Leo: It's a LIFO cube. Just have them pop off the ends. No problem. |
| Steve: And the other feature, Leo? |
| Leo: Yes. |
| Steve: Access to your logins on any device. |
| Leo: Any device? |
| Steve: They said. They said: "Move beyond single-device access and seamlessly sync and access your vault on any device, browser, or operating system." Leo, this is a breakthrough. Oh. |
| Leo: In their defense, it's a trial version. Right? I mean, they certainly don't expect anybody to use this all forever. |
| Steve: How could you use it for more than a day? |
| Leo: Yes, I know, but... |
| Steve: I don't get it. It must do some fancy stuff that I can't even imagine because, you know, I've seen some really ugly cars on the road, Leo. |
| Leo: Yeah? |
| Steve: Which says, if you make it... |
| Leo: We'll buy it. |
| Steve: ...somebody will buy it. |
| Leo: I know what car you're talking about, by the way. |
| Steve: Ugh. |
| Leo: It's really ugly. |
| Steve: Oh, boy. |
| Leo: And we're not talking the Aztek. |
| Steve: Especially when you go crazy with a paint job. |
| Leo: Oh, my, some of those creative wraps, yes. |
| Steve: They also, this Dashlane, also, if you upgrade, you get real-time phishing protection. |
| Leo: Real-time? |
| Steve: Yeah. |
| Leo: Not later? |
| Steve: Right then while you're logging on, Leo. It doesn't come back and say, oh, by the way, you got hacked last week. They say: "Set up your security and stay ahead of AI-powered" - you know, because AI, that's, you know, it's everything's AI powered now - "AI powered phishing with real-time alerts that warn you before you autofill your info on a suspicious site." |
| Leo: Warning. Warning. |
| Steve: Yes. But, you know, password managers won't fill in your info on a phishing site. |
| Leo: Because it's not the real site. |
| Steve: Because the URL won't match. So anyway, as a consequence of what I learned, I no longer imagine that any, not one, of our users could possibly, possibly be using the free Dashlane product. So I see no need to warn of the impending end of their free tier. The only other thing I learned is what a great deal 1Password and Bitwarden offer. Because, guess what, folks. You get all that for free. |
| Leo: I don't know what the 1Password free tier does, but I do - and I do remember this happened with LastPass, remember, they killed their free tier, and it upset people quite a bit. And so when Bitwarden said, you know, we would like to buy ads, one of the first questions I said is, so you have a free tier, unlimited passwords, unlimited passkeys, hardware keys, unlimited devices, all that stuff. But how long? Are you going to ever kill that? And they said no. We will never kill that. We can't. We're open source. Somebody would just fork it, and then it would continue on. So even if we wanted to, we can't. And that was actually one of the many, but one of the criteria I used to decide whether we would do their ads. |
| Steve: That's, I mean, that's an inarguable point, too. I mean... |
| Leo: Yeah. They can't. |
| Steve: Yeah. |
| Leo: They can't take it away. |
| Steve: Yeah. So if anybody out there is paying because you want - I mean, I pay my $10 a year to Bitwarden. |
| Leo: A year. And you don't have to, by the way. |
| Steve: No. |
| Leo: It's free free. |
| Steve: But I want to support them. |
| Leo: Yeah, we want to support them, yeah. |
| Steve: Yeah, but wow. I just - and there's a lot of people apparently using that. |
| Leo: To be fair, Dashlane is a perfectly good password manager. |
| Steve: I don't want to be fair. |
| Leo: It's just not free. Just don't - there's no free tier now. |
| Steve: Like why even - yeah, yeah. |
| Leo: But, I mean, it's not a bad - I've used Dashlane. It's fine. I mean, I've tried them all. I have to, to kind of keep my eye on what's going on. It's pretty. |
| Steve: So it's not Crashlane or Trashlane. |
| Leo: It's not Crashlane. Although one of our Discover members said, "Oh, don't use Dashlane. Use Dashhighway. It's got many more free passwords." You're on the lane version. That's the problem. |
| Steve: Yes, that's right. Wow. |
| Leo: No, that's one of the things I'm very proud of. We very carefully vet all of our sponsors, and we make sure that, for instance - the problem is you can't guarantee a company won't be acquired, and that's what happened to LastPass. We vetted them. |
| Steve: I was going to say. And you can't guarantee something won't happen to them. |
| Leo: Right. |
| Steve: You know, we're a little sorry that in the past we were all pro-LastPass. But in the past they were doing the right thing. |
| Leo: Right. And it's one of the reasons that I support open source. Because, as you see, even if you decided, you know, you were going to withdraw that free product, it's open source. Somebody just forks it and says, well, Leo's version is going to be free forever. And that's that. |
| Steve: Pretty much. |
| Leo: Right, yeah. I like open source. I'm just looking to see, I don't know if there's a free tier on 1Password. I think companies probably shouldn't really offer the free tier. Apparently that's a problem. |
| Steve: I do understand asking for a reasonable fee in return for all these services. |
| Leo: Yeah, yeah. |
| Steve: So, yeah, I just couldn't - I couldn't believe that anybody - my main point was how could anybody use a password manager with a 25-password limit? I just - I don't get it. |
| Leo: Oh, let me see. Get started free. I'm looking at 1Password. But it's basically a free trial that you can have for 14 days, and then you either it kicks in or it doesn't. Yeah, I understand offering a trial version so people can see if they, you know, like the user interface or whatever. |
| Steve: Yeah. And word of mouth probably is like, somebody else is using Dashlane and says, oh, you've got to try it, it's the greatest. So someone says, okay, I'll try it. And they, you know, put a few passwords in and see that it works. And it's like, okay. |
| Leo: Eh, you put too many in. Just keep using those same 25 over and over. |
| Steve: I just can't imagine. You add a password, and it says, oh, sorry, you've reached your limit. I just - that's - I can't believe it. |
| Leo: Well, and your point is also a technical point, which is it doesn't cost them any more if you have 26 passwords than if you have 25. Maybe if you had 25,000. Anyway... |
| Steve: Break time, and then we're going to talk about more cool stuff. Okay. So last Wednesday, Socket Security detailed two particularly nefarious NPM packages they discovered in the repository. Their article was titled "Malicious npm Packages Target WhatsApp Developers with Remote Kill Switch," with the teaser "Two npm packages masquerading as WhatsApp developer libraries include a kill switch that deletes all files if the phone number is not whitelisted." Okay. So here's the top of their posting. They said: "Socket's Threat Research Team discovered two malicious npm packages specifically targeting developers building WhatsApp API integrations with a remote-controlled destruction mechanism. Published by npm user nayflore using email idzzcch@gmail.com, both naya-flore and nvlore-hsc" - those are the two package names - "masquerade as WhatsApp socket libraries while implementing a phone number-based kill switch that can remotely wipe developers' systems. The packages have accumulated" - and here's the worrisome thing - "over 1,110 downloads in a month..." |
| Leo: What? |
| Steve: Yeah. So there's 1,110 unwitting developers who downloaded this thing, thinking, hey, great, I need an API library for my enterprise's WhatsApp development, and run the risk. Or maybe they did get wiped. And as of this reporting, it still remained active on the npm registry. Socket said: "We have submitted takedown requests to the npm security team and petitioned for the suspension of the associated account." Good. They wrote: "WhatsApp Business API adoption" - and that's where this is used. "WhatsApp Business API adoption has surged, with over 200 million businesses now using the WhatsApp platform globally. This growth has created a thriving ecosystem of third-party libraries and tools for WhatsApp automation. Developers regularly install packages like whatsapp-web.js, baileys, and similar libraries to build chatbots, customer service automation, and messaging integrations. The packages published by nayflore" - this bad guy - "exploit this trust by positioning themselves as alternative WhatsApp socket implementations. "The malicious packages first retrieve a remote database of whitelisted, basically good phone numbers from a GitHub repository. Both packages use Base64 encoding to obfuscate the endpoint URL." And in their documentation in the show notes I have the URL for anyone who's interested. It's a JSON file. They wrote: "The Base64 encoding conceals the GitHub endpoint from casual inspection. The database is hosted on GitHub Pages, making it appear legitimate while providing the threat actor with remote control over which phone numbers will trigger destruction. "The malicious kill switch logic is embedded within the requestPairingCode function, which developers would naturally call when setting up WhatsApp bot authentication. This function appears legitimate and necessary for WhatsApp integration. When requestPairingCode executes, it immediately begins the kill switch process. The logic is simple: If the phone number exists in the remote database, the package continues normal operation. If not found, the function sets the getsNumberCode to '0000' and executes" - and you'll get a kick out of this command, Leo, because I have it in the show notes - "and executes 'rm -rf *'..." |
| Leo: Oh. Oh, no. Oh, no. Oh, no. |
| Steve: "Rm" of course is short for "remove." And they wrote: "...which recursively deletes all files in the current directory." |
| Leo: Yeah. |
| Steve: You don't want that executed on your machine. |
| Leo: No. The "f" is for "forced." The "r" is recursive. So basically... |
| Steve: Yup, recursive force star. |
| Leo: ...delete everything. |
| Steve: Exactly. |
| Leo: Wow. |
| Steve: They said: "This approach allows the threat actor to maintain a whitelist of 'safe' phone numbers." And those may also - those are probably those of known security researchers, or maybe their own people, although I would never be comfortable, you know, like using a "if this number is in the list, then don't delete everything." That seems crazy. But, I mean, I get it. Maybe they just would never install this anywhere, like on their own stuff. I mean, the whole thing seems a little screwy to me. But this is what the thing does. They said: "The pairKey parameter is particularly clever. It makes the function signature look more authentic for WhatsApp development while having no impact on whether your system gets destroyed. It's a clever social engineering touch to make developers think this is a legitimate WhatsApp pairing function." So the point is it goes through the motions that you would expect from a WhatsApp API tool, all the while if your phone number that you're pairing, that you're developing with, is not in this whitelist, that command is going to get executed on your system. They said: "Both apps contain identical generateCreds functions capable of exfiltrating device information to, and then they give the URL, api[.]verylinh[.]my[.]id/running, but the calls to this function are commented out in both packages. This suggests the threat actor initially planned data exfiltration and collection, but simplified the attack to focus purely on destruction. And since the kill switch executes rm -rf * immediately when a phone number is not whitelisted, any subsequent exfiltration attempt would fail on the destroyed system." Correct, because there's nothing left to exfiltrate. They said: "For whitelisted numbers that continue normally, there's no valuable data to steal anyway." So they finish, saying: "The presence of complete, functional exfiltration code indicates the threat actor has the infrastructure ready and could easily reactivate data collection in future versions by simply uncommenting the function calls." Anyway, but our takeaway here, I mean, the thing to get from this is for all developers to be deadly serious about maintaining some sort, whatever sort you want, of incremental versioning backup which will not delete any previously backed-up files, even in the event that a file is deleted from the backup source. As our listeners know, for many years I was using and loving the free Sync.com encrypted file backup and cross-machine synchronization system. But they went through a service outage rough patch a couple of years ago, and the inconvenience of that was enough to kick me back over to Syncthing. And Leo, you and I are both huge Syncthing advocates. |
| Leo: Huge, huge, yeah. |
| Steve: It is truly terrific. But I have not been as big a fan of its versioning, whereas Sync.com really has that nailed. So, being something of a belt and suspenders guy, I'm now using both. If anyone is interested, I wanted to revisit my endorsement of Sync.com. If anyone is interested in looking at it, you can use my referral code, and we each get an extra gigabyte of storage. I've already got plenty, but it's worth getting one. That means you start out with 6GB of free - and it's free and unlimited - to play with rather than their 5GB free plan. And the referral code hasn't changed from years ago, it's grc.sc, it's one of our shortcuts, grc.sc/sync. And that'll just bounce you over to the ability to set up a free account with an extra gig. |
| Leo: What don't you like about Syncthing's - I use the staggered versioning. |
| Steve: Yeah. |
| Leo: Which keeps, for me, you can have more, a year's worth of versions. |
| Steve: Yeah, I don't remember now. Maybe it was just to have - wanting a whole secondary, you know... |
| Leo: Yeah, maybe, yeah. Belt and suspenders, yeah. |
| Steve: Yeah, yeah. I guess I just - I sort of like browsing through the display. I was doing it for a while. I don't remember what it was that - anyway, I think there was - it was just it's nice to have both. |
| Leo: Well, you know what I do is I do a Syncthing to my NAS, which runs Syncthing locally. But I have it set for copy only. So even if you delete a file, it doesn't sync it. And that's what people get in trouble with. |
| Steve: Right. Right. You definitely, definitely don't - for example, if this rm, you know, recursive force everything instruction ran on your system, you don't want anything that looks at your system going, oh. |
| Leo: Oh, you deleted that. |
| Steve: You need to get rid of that, yeah. |
| Leo: Yeah. Because that's exactly what these guys are going to do to you. Incidentally, it strikes me that the big threat with these kinds of hacks - whoops. It keeps turning that on. I don't want it turned on. These hacked libraries is so many vibe coders now will be not really knowing what they're doing. |
| Steve: They'll be just sucking it in. |
| Leo: Oh, look, the code runs great. And they're pulling in these libraries, and they don't really know what these libraries are doing. It seems like this is going to be more of a risk. |
| Steve: Essentially they're taking a whole bunch of building blocks and plugging them together. |
| Leo: Right. And they don't really code in many cases. They don't know what's going on, yeah. |
| Steve: Right, yeah. |
| Leo: Hey, by the way, congratulations to Grayson Petty. He just donated $10 as a YouTube viewer. He says he just passed his CompTIA CySA+ exam, a score of 770, got all four CompTIA certifications: A+, Network+, Security+, and CISA. And that's because he listens to this show. So good job. He's one of our superfans in the chat. |
| Steve: Very cool. Very cool. |
| Leo: Yeah, congratulations, Grayson. Thank you for this. |
| Steve: Well, while we're on the subject, Socket also found 11 malicious Go libraries that download and run malware on infected systems. And GitLab's own security team found five malicious PyPI packages targeting the dev ecosystem of the Bittensor crypto wallet. So, you know, as they say, there's no such thing as a free lunch. It is so sad, Leo, that the beautiful concept of a community of well-meaning open source developers and contributors who want to help each other, that it's just been infected by miscreants. |
| Leo: Miscreants. Yes. Youts, clearly. |
| Steve: I do have some happy news. |
| Leo: Oh, good. |
| Steve: At last, uBlock Origin's stalwart developer, Raymond Hill... |
| Leo: Gorhill. |
| Steve: Well, Gorhill... |
| Leo: Yeah, no, it's the same guy. Yeah, yeah. |
| Steve: Yeah, yeah. Gorhill. |
| Leo: His handle is Gorhill, yeah. |
| Steve: Yes, has just released uBlock Origin Lite for Safari. |
| Leo: Aha. |
| Steve: So for iPad, iPhone and Mac users. And since finding things within the Apple App Store has always been mysteriously difficult, I don't understand... |
| Leo: Their search is terrible. |
| Steve: It's just unbelievably bad, Leo. |
| Leo: So bad. |
| Steve: You put in exactly what you want, and it offers you 12 other things until you get to the thing that's named what you searched for. |
| Leo: Right. And often, by the way, shows you bad copies or, you know, kind of skeezy versions of the same thing. |
| Steve: Yes. |
| Leo: It's really annoying. |
| Steve: I don't get it. Anyway, I have a link to the exact uBlock Origin Lite app on the Apple.com, you know, App Store in the show notes. |
| Leo: Good. |
| Steve: So if you don't want to dig around through it, you can find it from me on the show notes. |
| Leo: I'm assuming that he did this because he had to write it for Chrome; right? |
| Steve: Correct. |
| Leo: To get past the version Manifest 3. |
| Steve: Yeah. And basically he developed all the "lite" technology in order to do as much as he could with the more of the hands-off approach which the Manifest V3 requires. So he figured, hey, I could do that on Safari now. |
| Leo: Because Safari also has a severe limitations on what extensions can do. |
| Steve: Right. But uBlock Origin at least is the most we can get. And, you know, it's still great. The app, it just appeared in the store so it only has 84 ratings when I looked yesterday. But it's holding at 4.8 out of 5. |
| Leo: That's good. All right. |
| Steve: It's like almost all fives. And there were like a couple ones because somebody said, well, this doesn't wash my car. So it's like, no, that's not what it does. Great app. I encountered the following brief news blurb about SonicWall. All that this little blurb said was: "SonicWall has told owners of Gen 7 firewalls to disable the device's SSL VPN feature due to a security risk. The company says it received reports of attacks against the devices over the past three days from at least three security firms." I think actually there were four. "According to Arctic Wolf, Google Mandiant, and Huntress Labs, attackers hacked SonicWall systems and then deployed ransomware." And if it's hacking an SSL VPN, that is not good. "SonicWall says it's investigating to see if the attacks used older bugs or a new zero-day exploit." Okay. So first of all, this would and should be terrifying to anyone whose enterprise is behind any late-model - that's Gen 7 or later - SonicWall firewall which offers remote SSL VPN access. Not just one, but three major high-reputation security firms independently determined that ransomware was being deployed within enterprise networks via an unknown penetration vulnerability in SonicWall's firewall. It doesn't get much worse than that. And as we know, the idea of a remotely exploitable zero-day vulnerability in an SSL VPN would, sadly, not itself be very surprising since we've seen exactly that a number of times before. So I went looking for SonicWall's statement about this to find out what was up. So here's the clarification that they provided. They said: "Following our earlier communication, we want to share an important update on our ongoing investigation into the recent cyber activity involving Gen 7 and newer firewalls with SSL VPN enabled. We now have high confidence that the recent SSL VPN activity is not connected to a zero-day vulnerability," meaning not something new. They said: "Instead, there's a significant correlation with threat activity related to CVE-2024 [meaning last year sometime] 40766, which was previously disclosed and documented in our public advisory," and then they gave a number, it's from 2024, advisory number 15. They said: "We are currently investigating fewer than 40" - four zero, so it's not nothing - "fewer than 40 incidents related to this cyber activity. Many of the incidents relate to migrations from Gen 6..." |
| Leo: Aha. |
| Steve: "...to Gen 7" - yup - "firewalls, where local user passwords were carried over during the migration and were not reset." |
| Leo: Whoops. |
| Steve: "Resetting passwords was a critical step outlined in the original advisory." So people didn't follow the instructions carefully. |
| Leo: Yeah, yeah. |
| Steve: They said: "SonicOS 7.3 has" - meaning the new, the current - "has additional protection against brute-force password and MFA attacks. Without these additional protections, password and MFA brute force attacks are more feasible." And, you know, this is the publisher saying that, so we would word that to say "Password and MFA brute-force attacks are entirely doable." |
| Leo: Yes. |
| Steve: Right. Sorry about that. |
| Leo: More feasible. |
| Steve: More feasible. Well, yes, and also entirely doable. Many people have succeeded, in other words. |
| Leo: Yes. |
| Steve: So they followed that with their updated guidance, and something there also stood out. They said: "To ensure full protection, we strongly urge all customers who have imported configurations from Gen 6 to newer firewalls to take the following steps immediately. First, update firmware to version 7.3.0 [the latest current] which includes enhanced protections against brute force attacks and additional MFA controls. Reset all local user accounts for any accounts with SSL VPN access, especially if they were carried over during migration from Gen 6 to Gen 7. Continue applying the previously recommended best practices. Enable Botnet Protection and Geo-IP Filtering. Remove unused or inactive user accounts. Enforce multifactor authentication and strong password policies." They said: "If any local administrator accounts have been compromised through the earlier CVE-2024-40766, attackers may exploit administrative features such as packet capture, debugging, logging, configuration backup, or multifactor authentication control to obtain additional credentials, monitor traffic, or weaken the overall security posture." And now we know that includes installing ransomware. They said: "It's advisable to review any packet captures, logs, MFA settings, and recent configuration changes for unusual activity, and rotate any credentials that may have been exposed. We appreciate the continued support from third-party researchers that have helped us throughout this process, including Arctic Wolf, Google Mandiant, Huntress" - oh, yeah, and the fourth one is Field Effect. Okay. So those of our listeners who have been following along at home may have noticed something in their list of their remediation and preventive measures that I don't think I've seen before. Their fourth bullet point was "Enable Botnet Protection and Geo-IP Filtering." And the Geo-IP Filtering phrase is key. It is, in their notes, it's a link. I clicked on the link. It links to a page describing what they mean by the term, where they wrote: "Geo-IP Filter allows administrators to block connections coming to or from a geographic location by resolving the Public IP address to a particular country. This feature is usable in two modes, blanket blocking or blocking through firewall access rules. "Blocking through firewall access rules gives a network administrator greater control over what traffic is and is not scanned by the Geo-IP Filter. This is useful for deployments in which Outbound traffic may want to be uninhibited, but Inbound traffic should be subject to scanning. Typical deployments of Geo-IP Filter with firewall access rules includes DDoS and other network attack mitigation, as well as anti-spoofing." So this is great to see. Naturally, everything we've seen and learned informs us that any and all such SSL VPNs, other things like public-facing web management portals, and anything similar should be locked down out of the box with the engineer who is configuring it forced to selectively enable only the country or countries from which valid remote access is expected to originate. Alas, the industry is not there yet, but at least we're seeing progress. Having SonicWall offering such a feature right there in its user interface at least means there's a chance that a security-oriented engineer who is offered the option may take the hint. And note that the beauty of an IP-based filter is that no one scanning the Internet from Russia or from China or anywhere else outside the allowed jurisdiction will detect that anything at all is there. They're not frustrated. They're not having their credentials denied. They don't have a page prompting them to log in and saying, sorry, that's not the right password, try again. Their scanning packets are simply dropped. They will never know anything is there for them to struggle to get. So, I mean, it is the way to do this. But no one's doing it yet. At least this is some forward motion, and it's good to see. Cisco's Talos security group headlined last week's disclosure: "ReVault! When your SoC turns against you." And this is what I was talking about when I talked about all the effort that security researchers need to go through unfortunately in order to help the companies who have security problems. Leo, let's take another break, since we're an hour in. |
| Leo: Yes, sir. |
| Steve: And we're going to talk about what Cisco and Talos found in Dell's laptops and why anybody with a Latitude needs to make sure they get this update. It's a complete security bypass that can be done from any app running on their laptop with no permissions. |
| Leo: All right. Now back to Steve. |
| Steve: Thank you, Leo. So Cisco's Talos Security Group headlined their disclosure "ReVault! When your SoC turns against you." |
| Leo: It's a revault, I get it. It's revaulting. |
| Steve: That's right. It's revaulting. SoC stands for System on a Chip, and the Vault in the name stems from a subsystem on the huge number of Dell laptop PCs, more than a hundred different models representing tens of millions of Dell laptops because it's the Latitude, which is their headline laptop. The subsystem is called the ControlVault3. So again, the short news blurb that caused me to look deeper just said: "A set of vulnerabilities can allow threat actors to take control of tens of millions of Dell laptops. The bugs impact the ControlVault3 firmware that's used to safely store passwords and biometric data inside a secure chip on Dell Windows laptops. The five bugs, codenamed ReVault, impact more than 100 Dell laptop models. The bugs can be exploited" - and this is where this got, it's like, whoa - "via a Windows API and don't require, or doesn't require, elevated privileges. Dell has released firmware updates." Okay. So by far the most worrisome part of the entire statement is: "The bugs can be exploited via a Windows API and don't require elevated privileges." Most flaws that we encounter in security device firmware are actually kind of obscure. You know, they require things like boot-time access, or access to the system motherboard's management interface or something. But here we have a set of flaws that literally ANY Windows app running on anyone's Dell laptop under their non-UAC minimal user privilege account could exploit. So here come the miscreants. As I mentioned, Cisco's Talos security group discovered and publicly disclosed this last week. I'm sure their discovery was much earlier, since they waited until Dell had created and tested and published the required firmware updates. The good news is those updates exist. The bad news is that they need to be installed before any of those more than 100 Dell laptop models representing tens of millions of physical laptops will be made safe. So let's see what Cisco's Talos group disclosed last week. Their report leads with four bullet points. First, Talos reported five vulnerabilities to Broadcom - it's actually a Broadcom chip in the Dell laptop - to Broadcom and Dell affecting both the ControlVault3 Firmware and its associated Windows APIs that we are calling ReVault. Second bullet point: 100+, meaning more than a hundred, models of Dell Laptops are affected by this vulnerability if left unpatched or, you know, until patched. Third, the ReVault attack can be used as a post-compromise persistence technique that can remain even across Windows reinstalls. So it's like a rootkit. Once it gets in there, I mean, worse than a rootkit. It's not even going away after a reinstall. And, finally, the ReVault attack can also be used as a physical compromise to bypass Windows Login and/or any local user to gain admin and system privileges. Yikes. They continued: "Dell ControlVault is a hardware-based security solution that provides a secure bank that stores" - as in storage bank - "that stores your passwords, biometric templates, and security codes within the firmware. A daughter board provides this functionality and provides these security features through firmware. Dell refers to the daughter board as a USH, a Unified Security Hub, as it's used as a hub to run ControlVault, connecting various security peripherals such as a fingerprint reader, smart card reader, and NFC reader." So, you know, that's an elegant design, to like have all of those physical devices run through a physically separate standalone board before they get to have any access to the motherboard. Good design. They wrote: "The current iterations of the product are called ControlVault3 and ControlVault3+ and can be found in more than 100 different models of actively-supported Dell laptops" - and I've got links for all this, by the way, at the end of this. They said: "...mostly from the business-centric Latitude and Precision series. These laptop models are widely used in the cybersecurity industry, in government settings, and in challenging environments in their Rugged version. Sensitive industries that require heightened security when logging in (via, for example, smartcard or NFC) are more likely to find ControlVault devices within their environment, as they are necessary to enable," that is, "ControlVault is necessary to enable these enhanced security features." So that's just great; right? It's the machines that are most in need of additional security and would be more likely to be targets, that have had their security dramatically impacted by the discovery of these bugs. That's a big whoops. They said: "Today, Talos is publishing five CVEs and their associated reports. The vulnerabilities include multiple out-of-bounds vulnerabilities" - in other words, good old buffer overflows - "an arbitrary free, and a stack-overflow, all affecting the CV firmware," you know, the ControlVault firmware. "We also reported an unsafe-deserialization" - you know, that'll be some interpretation of something - "that affects ControlVault's Windows API." So there was a bug in the Windows code and four bugs in the ControlVault firmware. So think about that. To accomplish this, Talos would have had to extract and reverse-engineer the proprietary firmware from the Broadcom chip. Dell certainly didn't say, "Hey, please check the firmware we wrote for our core security chip which provides all of the most critical physical and biometric security for our most secure laptops." My point is here, it's really a shame that this sort of symbiotic relationship doesn't, you know, that that sort of symbiotic relationship doesn't exist between manufacturers and security researchers, where manufacturers could be of more help to researchers. How many times have we looked at all the extra and unnecessary effort security researchers have had to go through just to reverse engineer and obtain the same information that the manufacturer already has sitting in a file somewhere? And after all that work, which might well come to nothing, right, they might have extracted the firmware, reverse engineered it, taken a good hard long look at it and found no problems. But even after they did, the security researchers say to the manufacturer, "Hey, we just worked our butts off thanklessly for several months to discover a set of five really horrendous security vulnerabilities that affect tens of millions of your most security-essential laptops." You know, it's not good news that they're providing, but it is potentially, you know, heading off a horrific exploit against all of those laptops. So there's something still very wrong with the way we're doing all of this today. The economics in our system are not producing the right incentives. Cisco's Talos group concluded, writing: "With a lack of common security mitigations and the combination of some of the vulnerabilities mentioned above, the impact of these findings" - that is, what they discovered, they're trying to give us some context here, what this means - "is significant." They said: "Let's highlight two of the most critical attack scenarios we have uncovered. First, post-compromise pivot." They said: "On the Windows side, a non-administrative user can interact with the ControlVault firmware using its associated APIs and trigger an Arbitrary Code Execution on the CV firmware." Okay, now, given what we know, it's likely possible for the user, meaning an app some unwitting user runs, to load an extra-large buffer of executable code into ControlVault's RAM and then cause that buffer to be executed, thus running their own CV firmware code on the ControlVault in order to get up to some mischief. Talos said: "From this vantage point, it becomes possible to leak key material essential to the security of the device, thus gaining the ability to permanently modify its firmware. This creates the risk of a so-called implant that could stay unnoticed in a laptop's CV firmware and eventually be used as a pivot back onto the system in the case of a Threat Actor's post-compromise strategy. We show how a tampered CV firmware can be used to hack Windows by leveraging the unsafe deserialization bug mentioned previously." And then that was just one. The second is the physical attack. They said: "A local attacker with physical access to a user's laptop can pry it open and directly access the USH board over USB with a custom connector." So there's a USB access to the chip. "From there, all the vulnerabilities described previously become in-scope for the attacker without requiring the ability to log into the system or knowing a full-disk encryption password. While chassis-intrusion can be detected, this is a feature that needs to be enabled beforehand to be effective at warning of a potential tampering." And it's typically not enabled by default on the BIOS. "Another interesting consequence of this scenario is that if a system is configured to be unlocked with the user's fingerprint, it is" - get this - "also possible to tamper with the CV firmware to accept any fingerprint rather than only a legitimate user's." So when you think about that, the "any fingerprint" attack is sort of diabolical. How often does anyone go around asking random people to verify that their fingerprint does NOT unlock their laptop? Probably not often. Perhaps never. The affected user would simply notice that their fingerprint reader had apparently suddenly become much better at accepting their fingerprint than it previously had been. Anyway, Cisco says: "To mitigate these attacks, Talos recommends the following: ' Keep your system up to date to ensure the latest firmware is installed. CV firmware can be automatically deployed via Windows Update, but new firmware usually gets released on the Dell website a few weeks prior.'" So that's good to know. That suggests that Windows, through Microsoft and through Dell through Microsoft, will be pushing this out through Windows Update. So it would be nice to have some way to verify that. I didn't dig into this to see whether that is available, but that's great. "If not using any of the security peripherals (the fingerprint reader, the smart card reader, and the NFC reader), it is possible to disable the CV services (using the Service Manager) and/or the CV device (using Device Manager)." And that's, of course, that's standard best security practice; right? If you don't need any of these devices, turn them off. Then you don't have to worry about them getting used and hacked behind your back. And finally, they said: " It's also worth disabling fingerprint login when risks are heightened, for example, leaving one's laptop unattended in a hotel room. Windows also provides Enhanced Sign-in Security (ESS), which may help mitigate some of the physical attacks and detect inappropriate CV firmware." Then they said: "To detect an attack, consider the following: Depending upon your laptop model, chassis intrusion detection can be enabled in the computer's BIOS. This would flag physical tampering and may require enabling a password to clear the alert and restart the computer. In the Windows logs, unexpected crashes of the Windows Biometric Service or the various Credential Vault services could be a sign of compromise. And Cisco customers using Cisco Secure Endpoint can be made aware of potential risks with the signature definition 'Loaded by Abnormal Process.'" Anyway, so Dell's own pages label this CRITICAL, in all caps, and they provide a 36MB, not very big, download to patch this. It's a Windows executable. For any Dell Latitude or Precision laptop owners who would like to be proactive and get this patched, I've got the links in the show notes. Hopefully, given that it's previously known, Dell has published it, where are we? Today is the second Tuesday of this month. I don't know which month, which Patch Tuesday this may have been pushed out. But maybe people are getting it today. Maybe they got it last month. I don't know where Cisco timed their release. So it might be that if you try to install the firmware, you'll be told it's already been updated, and you don't need to install. But anyway, anybody with a Dell laptop, a Dell Latitude, probably would be well served to take a look at this and see about getting this resolved. Make sure that they do have this latest firmware. Okay. So the listener of ours, Rosco, who I mentioned briefly last week, who is in the field, is responsible for managing and maintaining Microsoft systems, wrote. He said: "Hi, Steve. With respect to the SharePoint on-prem patching issue, it's important to understand that the ecosystem can be highly complex, and patching can be more difficult than it seems. Office 365 might seem to be an obvious way to resolve all these issues, but it can be problematic, too. "The Enterprise office suite has many components, which form an extensive requirements matrix and consists of Windows server version, Active Directory version, Exchange Server version, Dynamics CRM Server version, SharePoint Server version, MS-Project Server version, Dynamics Great Plains accounting, and Dynamics Human Resources," he said, "to name some of the commonly deployed solutions, although some have been withdrawn as on-prem installable components." Again, Leo, as you noted, because Microsoft is trying to move everybody up to the cloud. He said: "The versions of all of these components need to be harmonized in order to have a viable, working installation. As a result, in order to update or patch to the latest SharePoint on-prem version, the trickle-down requirements might extend to updating and patching any or all of the other components in the service stack. In extreme situations, this can result in days or even weeks of applying patches, backing out, applying patches in version order or to different services first, for example, applying a patch to Active Directory first, then to CRM, then finally to SharePoint, rather than to SharePoint first. The result can be almost unmaintainable, especially for a small-to-medium-enterprise with limited IT resources. "Here's an entirely plausible cascade that demonstrates the deep interconnectedness: So the sun is shining, the birds are singing, the grass is green, systems are stable, and everything is beautiful in the world," he writes. |
| Leo: Woohoo. |
| Steve: Yes. "A critical vulnerability is discovered in SharePoint with CVSS 9.8, and a patch is available." Well, we know that's exactly what happened. He says: "The installed SharePoint version is two patch rollups behind, so SharePoint has to be brought up-to date. The second SharePoint rollup will not run on the currently installed Active Directory, so an Active Directory upgrade is required. The Active Directory upgrade implies a Windows server upgrade. The new Active Directory version no longer supports the installed Exchange version, which also must be upgraded. The new Active Directory version also deprecates NT LanMan authentication, which Microsoft SQL was still using, so Microsoft SQL is also in scope. "The Exchange end of the Dynamics CRM Exchange API deprecates two methods used for email integration for mail-outs to customers and reception of replies in order for replies to be tagged to the original outgoing message inside CRM. Thus CRM is now in scope for two rollup installations. The second of these CRM rollups deprecates an API method being used for integration with the parent company's reporting tools which are required to report the subsidiary's sales pipeline prospects to the relevant stock exchange, which is a legislated requirement. "The parent company must upgrade their data interchange tooling." Alarm bells. "The parent company runs Oracle EBS. Seasoned operators might guess which freight train is heading down the tracks. Installing the rollups in CRM also breaks three in-house customizations, which must be redesigned and reimplemented. At long last, the full cascading set of upgrades has been deployed, and the SP rollups and patch can be installed." He says: "This is the sort of thing that actually happens, which can result in weeks" - no kidding - "of disruption to business activities and manual workarounds." Is it any surprise, Leo, that everyone just wants to leave everything the way it is? |
| Leo: Yeah. If it works, don't fix. |
| Steve: Holy crap. |
| Leo: Don't fix it if it don't break, or something like that. |
| Steve: Yes. And unfortunately you can't just patch the one problem. |
| Leo: Right. |
| Steve: You know, you get a rollup of all these things which have a network of interdependencies with other versions of things because Microsoft is always keeping everything up to date because they don't have all of the connections to external third-party tools or to internal things that have, you know, in-house customizations which all would have to be moving forward. And Microsoft is adding and removing APIs that break things as we move forward. It is just a mess. As Rosco wrote: "The end effect can be an erosion of trust in the technical solutions provided by Office 365 and these bolt-on components, hence Microsoft has pushed clients towards a cloud-based subscription model which provides overall greater stability." In other words, because Microsoft, because of like the way this ecosystem developed, it just became untenable. |
| Leo: Right. It's a house of cards, yeah. |
| Steve: Yes, from a maintenance standpoint. |
| Leo: This is why I have such respect for IT professionals. This is a miserable job, yeah. |
| Steve: Yes. Yes. He said: "Organizations may choose to wait for patch rollups in order to reduce the inevitable in-depth troubleshooting that can occur if the application of patches results in unexpected behaviors of seemingly-unrelated software components, for example, a patch applied to SharePoint causing unexpected behaviors in CRM." He said: "On-premises SharePoint offers a broader feature set, however" - and this is the problem. "On-premises SharePoint offers a broader feature set, such as the ability to design and deploy business processes directly into SharePoint, using various third-party business process management add-on tools and database connectivity and functionality required to support business processes." Meaning you can't do that once you move to the cloud. So you lose functionality when Microsoft says, oh, we want you to go to the cloud. |
| Leo: Right. |
| Steve: He said: "Online SharePoint, however, reduces support for business processes to Microsoft's own development tools, logic apps, flows, web jobs, and functions, implying that an organization's significant investment in third-party tools and business process deployment and testing becomes redundant, and the organization is forced to either adopt a completely new solution outside of SharePoint, or reengineer all their existing business processes to conform to Microsoft's provided frameworks, at possible significant cost and the possibility of losing significant features and functionality." He said: "In corporate environments, such a significant change can mean considerable embarrassment to decision-makers who advocated for the third-party development approach in the first place, which can add to inertia in moving to a cloud-based solution." |
| Leo: Yeah, but you have to understand that that's why it's so complex to update it. |
| Steve: Yes. |
| Leo: Because you have this complex ecosystem. |
| Steve: Yup. |
| Leo: So you can't complain about how hard it is to update it if you've opted for this crazy quilt of capabilities, either. |
| Steve: And you've taken advantage of them all. |
| Leo: Right. Yeah. |
| Steve: Anyway, his perspective leaves me with a much deeper appreciation for the fact that, in my own little tiny world, I have never needed to become embroiled in any of that. What a monumental mess. |
| Leo: When Richard Campbell talks about running his own Exchange and SharePoint server out of the house, I just tip my hat to him, is all I can say. You're a better man than I. |
| Steve: You know, it feels as though the utter lock-in which has resulted, it was incremental. And it was more or less inadvertent. I mean, but, you know, today's Microsoft is also well aware of the fact that businesses have gone all-in to their solutions, and that those that have no longer have any meaningful way out. They have no freedom. They are, you know, they're being pushed to migrate to the cloud. It's the only way to stay secure, basically transfer responsibility to Microsoft; hope that Microsoft doesn't break their stuff because now it's no longer under their control; and pay a subscription on all of this stuff, which they also have no control over. Microsoft can charge anything they want because it is a captive audience now. Wow. Okay. |
| Leo: And you know, if Microsoft's going to take this on, let them. Good for them. Better them than us. |
| Steve: So a listener of ours, Brian Savacool, said: "Steve, thanks for all that you do. You had recommended a picture tool for Windows that could correct the Keystone or perspective of a picture-of-the-day image which was severely skewed. I can't find it in any of the emails or show notes from the past year. Could you please bring it up in the next feedback section? I'm writing some user-documentation, and the sample image needs similar correction. Thanks for your weekly breakdowns. It really helps me at the help desk where I work. Brian." So I received a number of inquiries about that, and I have been remiss in not replying to them all. So I wanted to use Brian's question from last week to get caught up. The tool is free. As I think I've mentioned before, it's funky and somewhat finicky. It's by a German guy, and he just sort of has his way of doing things. So it's not perfect, polished, and proper. But it's the tool I use because the German guy who created it got the basic mechanics exactly right. When I first mentioned this on the podcast, I received a ton of feedback from our listeners about alternative solutions. The one that sticks in my mind was someone commenting that he recalled that my own go-to graphic editing tool, which is PaintShop Pro, and he's right, has built-in perspective correction. And again, he's right, it does. But it's the sort of perspective correction that iOS has, where you have dials for horizontal or vertical distortion. The thing that the - and this is the name of it, it's just called Perspective Image Correction. The thing the Perspective Image Correction app got exactly right is that its operator rubber bands the vertices of a four-sided box whose opposing sides should be parallel - yup, you found it - in the final image, and it then does that. To my way of thinking it's really the correct way to solve the problem. The app lives over on SourceForge. I have a link to it in the show notes, and I also created a GRC shortcut for anybody who wants to find it. Not surprisingly it's grc.sc/perspective. And that will bounce you over to the download page on SourceForge. And it's just a Windows app. You know, it may not be the perfect solution for everyone, but it's the best one I have found, and the price is right. It's free. And it stopped me looking for anything better. So there you go, Brian. That is the solution. Oh, and he also asked, actually it was in another email, he said: "Hi, Steve. You had casually mentioned on Episode 1035 about running SpinRite on a Kindle device. Can you explain in a future episode how that's accomplished? Is it done in the native boot environment, or does it require VirtualBox and a special device driver?" He said: "I have a 'Black Friday' special Android tablet that has been getting very slow and sludgy, even after multiple factory resets. I would like to run SpinRite on it before I give up and toss the thing. Thanks for any tips you may have." And he said: "I can't wait to try out the DNS Benchmark Pro once it gets released. All the very best, Brian in Schenectady." |
| Leo: Schenectady? |
| Steve: Schenectady? Is that a... |
| Leo: It's Schenectady. |
| Steve: Schenectady. |
| Leo: I'm not laughing. I'm not. I'm really not. |
| Steve: Schenectady, New York; right? |
| Leo: Schenectady, yes. But no shna. |
| Steve: No shna. |
| Leo: The shna is silent. |
| Steve: Schenectady. Anyway, so regarding the Benchmark, I wanted to mention that I'm very nearly finished with all of the new features for the base-model DNS Benchmark 2.0. It's finally working very nicely, and through this work I've obtained some clarification about the base model versus the Pro edition. The new v2.0 non-Pro edition will be full-featured, and will basically run on demand as a stand-alone app for Windows and WINE. So it does also run under Linux and Mac. Running on demand makes it useful for obtaining an immediate snapshot of any collection of up to 500 - way more than you'll need - remote DNS resolvers over IPv4, IPv6, DoH, or DoT, and all at the same time or individually, whatever you want, from the user's location. What we've seen is, just like the old saying about retail, it's "location, location, location." Our testers are spread around the globe, and it is surprising to see how much people's location changes their results. You know, not surprisingly. The DNS4EU resolvers, as I've mentioned, as incredibly slow for me over here in the states, but testers over in the EU have said they're very quick, among their fastest. One of the things we've also seen is that the time of day and the day of week also affects the Benchmark's immediate short-term outcome. If I run the Benchmark in the middle of a weekday, I consistently see a different result from running it on a weekend morning, for example. The difference may not be significant, but you won't know until you try it. So if all you had was the interactive edition, you might want to run it at the same time as you are using your machine at that location. And running it at different times can be useful. One very cool new feature is that the Benchmark is now aware of whether differences in performance are statistically significant. For example, two different resolvers might have slightly different average performance, but each resolver's individual spread of performance might be wide enough that it's not possible to say with 95% certainty that the differences seen were not just the result of random variations in packet transit times because those are varying with the Internet's conditions. So the Benchmark is now aware of the nature of this, and it incorporates this awareness into every conclusion that it reaches. The base model of the Benchmark tests each resolver against the Internet's top 50 domains three different ways. So it issues 150 DNS queries to each of the DNS resolvers being benchmarked. This requires about four minutes for 120 DNS resolvers, which is about all the time that an interactive benchmark should consume without making its user impatient to get some results. During this time, it's measuring the precise time taken for 150 different DNS queries to each of the resolvers that it's benchmarking. Now, it turns out that statistics can be annoying because, even with 150 individual timing samples per resolver, the individual variation among samples means that our ability to draw firm statistically significant conclusions remains somewhat limited. And that's where the Pro edition comes into its own because Pro will operate as a Windows service entirely in the background. Its user won't be sitting around impatiently waiting for results. Pro builds and maintains a database which will allow it to measure resolver performance across a continually broadening time horizon. If the user turns their machine on and off, it will automatically be measuring the times when they are using their machine. And if their habit is to leave the machine on, it will be aware of when their machine is unattended and note whether they're actually using it or not so that it's able to bias its statistics for when they're actually using the machine because of course that's what matters the most for DNS performance. DNS queries are very small and very lightweight, so they will never interfere with the normal foreground operation of the machine. And anytime a Pro owner wishes to see what's up, they simply launch the same Benchmark utility. Then they're able to choose whether they want to run an interactive benchmark on the fly right then, or if they're interested in browsing the aggregated data that the Pro version has been collecting in the background since they installed it. So of course, as usual, I have no idea when any of this will be ready. I actually think, Leo, I might have it ready before Andy has his website online. So I am getting close. But anyway, I'm becoming excited. The non-Pro features, basically all the main features of the Benchmark are fleshed out and finished and working in testing. And very soon I'll switch over to adding the background database aggregation and display, which, you know, I get to reuse the same UI, so that won't be a big deal. As for Brian's question about using SpinRite to revive an old Android tablet, until we get to SpinRite 7, which will be a pure Windows app, it is necessary to boot any system into DOS where SpinRite is able to run. For people who've never done this, I created that BootAble freeware which uses all of the same processes as SpinRite to create and prepare a bootable USB thumb drive. And you could just download that for free to play with it. If you're able to boot a PC with BootAble, then SpinRite will also work in the same way. So the only requirement for restoring the performance of any device's internal FLASH storage, whether it's a Kindle or an Android tablet, is being able to expose that internal storage as a drive. That might mean switching its storage mode. Android devices usually have a - it's sometimes called target drive mode, or USB target mode, something like that. The idea being that if you can attach it to a PC and have it in a mode where you see the Android's drive as a disk, then you can attach it to a PC and run SpinRite on it. |
| Leo: As a USB drive, basically. |
| Steve: Yes, yes, exactly. And you want to run a Level 3, which rewrites the drive's entire surface to restore its original factory performance. And of course all of this will be much easier once I get SpinRite moved over into Windows. But I have a few other things I want to get finished first. So at that point, once that's done, I can start that joyous project. I'll get there. I'll get there. |
| Leo: There's always something, isn't there. |
| Steve: Yeah. I love it. |
| Leo: Could be worse. You could be running an Exchange Server, so just be glad. |
| Steve: Yeah. I'm running hMailServer, and it just sits there and works. No problem. |
| Leo: Much easier. Much easier, yeah. |
| Steve: Listener Michael Swanson said: "Hi, Steve. I have had InControl running since" - GRC's freeware InControl - "running since shortly after you released it to ensure my laptop is not accidentally updated to Windows 11. However, KB5001716 did NOT install on my laptop until I released control. I believe KB5001716 is the update that provides the free Windows 10 security patch extension, though I still have not seen that offer pop up in the Windows Update dialog. Best regards, Mike." Okay. So first of all, Michael, thank you. It's interesting that InControl appeared to block the installation (or perhaps the re-installation) of KB5001716 since we know that InControl is designed to only block major version changes. So here's what Microsoft has to say about this update, this mysterious update. They said: "After this update is installed, Windows may periodically display a notification informing you of problems that may prevent Windows Update from keeping your device up-to-date and protected against current threats. "For example, you may see a notification informing you that your device is currently running a version of Windows that has reached the end of its support lifecycle, or that your device does not meet the minimum hardware requirements for the currently installed version of Windows." That's interesting. Does not meet the minimum hardware requirements. So it sounds like they're saying those people that have put Windows 11 on a machine that isn't supposed to have Windows 11. |
| Leo: Yeah, doesn't have [crosstalk]. |
| Steve: It's going to start complaining about that. |
| Leo: Eighth generation Intel chip, yeah. |
| Steve: So of course we saw this sort of thing, right, a lot during the forced migration to Windows 10, which is where I first created the app... |
| Leo: Never11. |
| Steve: Yes. No, it was called Never10. |
| Leo: Never10, I mean, yes. |
| Steve: Yes, Never10, where Microsoft would be updating their Windows Update system to introduce constantly evolving new messages about the coming end of the world. You know, it was almost comical. |
| Leo: They [crosstalk] down, by the way, with Windows 10. |
| Steve: Yeah. The dialogs back then, from people who had Windows 7 or 8, they gradually changed from, you know, would you like to update to Windows 10, and you had a choice of "No thanks" or "Yes please." And then after a while it changed to "now" or "tonight." So it's like, wait a minute. What happened to "never"? Thank you. Anyway, after receiving Michael's note, I just checked, and my Win10 21H1 was locked with InControl. When I go to Windows Update I see the expected red notification. It says "Some settings are managed by your organization." And then if I click on "View configured update policies," I see that the policies set on this machine are "Target release version for feature updates" and "Target product version for feature updates." And those are the things that InControl sets. So it looks like it's putting an end-user's machine under "organizational management" so that it won't do something that the IT department doesn't want to have it do. However, I just tried what Michael talked about, and I can confirm that whatever KB5001716 is, briefly releasing InControl and performing a manual Windows Update DID install KB5001716 into my normally InControl-locked Windows 10 machine. So I confirm what Michael reported. I don't know that I want 5001716. I don't know if I'm going to start betting bugged by Microsoft now. But what's interesting is this is not the first time that KB5001716 has been offered. More than a year ago, back on March 7th, so almost a year and a half ago, of 2024, Ghacks.net's Martin Brinkmann wrote about this under the headline: "Microsoft's sneaky KB5001716 Windows 10 update pushes Windows 11." He wrote: "If you run Microsoft's Windows 10 operating system on your devices and want to keep it that way, you may want to check whether the Windows 10 update KB5001716 is installed on the device. The reason for this is that it is designed to push newer versions of Windows, including Windows 11, to the device. Microsoft installs the update automatically on non-managed Windows 10 devices that have automatic updates configured." So of course InControl makes it look like a managed device, thus this isn't installed. So enterprise systems won't get this because they're managed, and under management is what InControl creates. So anyway, I've got some more about this in the show notes. But, you know, that's what I know at this point is that there is - basically this KB5001716 is a rolling update which is generically about Microsoft and notices and pushing you to newer versions and complaining when you haven't gone. And apparently now complaining when you have gone and you shouldn't have gone because you've got a chip that you're not supposed to actually be running Windows 11 on. Wow. So I don't think it's a problem that InControl is blocking it, but I wanted to let our listeners know, if they want that, just turn off InControl, manually run Windows Update, and you'll see - because I did. I ran Windows Update first to make sure that I was all caught up, and I don't think anything happened. And then I turned off InControl, immediately ran Windows Update again, and I saw exactly that KB5001716 installed itself. And then I turned InControl back on again. So I've got that, whatever that is. You know, I don't know if I care. But, you know, what the heck. Okay. A little bit of sci-fi news because something's happening tonight. I finished, I should mention, Andy Weir's second book, "Artemis." |
| Leo: Oh, yes. You like the ending, or no? |
| Steve: Actually, it came more quickly than I expected. |
| Leo: Yeah. And kind of abruptly, yeah. |
| Steve: Well, my Kindle showed me that I was 90% of the way finished with the book when the plot wrapped up and the story ended. |
| Leo: All done. Bye-bye. |
| Steve: And it turned out that the remaining 10% of the book was a kind of interesting Q&A discussion with Andy, and a discussion of the science and the underlying economic principles around which he had designed and created the scientifically and economically accurate, although entirely fictional, storyline. |
| Leo: I thought that was interesting, what it would take to have a colony on the Moon. I thought that was very interesting. |
| Steve: Yeah, yeah. And he really did dig into that. |
| Leo: Oh, yeah. |
| Steve: Again, it is science, and it is fiction. So if somebody wants, you know, like a really solid piece of work, that's what you get every time from Andy. And he did deliver it. So, yeah, it was an interesting story. I have to say, though, I'm now reading something wacky, like some scavengers have stumbled onto an alien ship of unknown origin that was trapped in a time bubble, and something has infected one of them, and it's like, really interesting. So I guess I kind of like other wacky sci-fi also. Anyway, shifting gears, I wanted to mention that today, August 12th, 2025, begins the continuation of one of our generation's major sci-fi franchises. Now, exactly two years before the debut of this franchise's first film, the world was taken by surprise on May 25th, 1977 with the release of a film that, surprisingly in retrospect, many people were unsure of. The movie's title was just "Star Wars," and needless to say the world changed that day. |
| Leo: Yes. |
| Steve: I still recall sitting in a large theater in Palo Alto, Northern California, with a big bucket of popcorn, having no idea what to expect, and being astonished. Then, weirdly, exactly two years later, by coincidence on the same calendar day, May 25th, but this time 1979, the world changed again when Ridley Scott directed an horrific alien creature from beyond our imagination to gestate inside an unwitting starship crew member, and to then explosively emerge through his chest. Like, just blew us all away. |
| Leo: Yeah, that's true. That was a great movie. |
| Steve: Oh my god. That scene and others were so over the top with surprise and tension that the movie "Alien" initially received somewhat mixed reviews. Reviewers, I think, were afraid to like it and weren't sure what to think. But the movie went on to win the Academy Award that year for Best Visual Effects - not surprisingly - three Saturn Awards for Best Science Fiction Film, Best Direction by Ridley Scott, and Best Supporting Actress, interestingly, not Sigourney, but Veronica Cartwright. |
| Leo: Oh. |
| Steve: She played the role of that kind of wimpy woman who was always screaming and freaks out a lot. |
| Leo: She was so scared, yeah. |
| Steve: She was really scared. And Sigourney's role didn't win her anything, but it definitely put her on the map. |
| Leo: Oh, yeah. |
| Steve: The film also took home a Hugo Award for Best Dramatic Presentation. Now, once the world had caught its breath, it was able to look at "Alien" more objectively. Wikipedia now writes: "In subsequent years, 'Alien' was critically reassessed and is now considered to be one of the greatest and most influential science fiction and horror films of all time." |
| Leo: Indeed. |
| Steve: They said: "In 2002, 'Alien' was deemed 'culturally, historically, or aesthetically significant' by the Library of Congress and was selected for preservation in the United States National Film Registry. In 2008, it was ranked by the American Film Institute as the seventh-best film in the science fiction genre, and as the 33rd-greatest film of all time by Empire." So for what it's worth, James Cameron's sequel "Aliens" is, without any doubt, one of my own personal favorite science fiction films of all time. For those who don't know, "Aliens'" original theatrical release was missing some very interesting additional plot development footage that was later included in various of the Director's Cuts of the movie. |
| Leo: Huh. Oh, I didn't know that. Huh. |
| Steve: They were worth watching. The scene where the family left the enclave and went out in their rover and actually brought the original infection back, that was completely missing from the original theatrical release. But we now know about that, thanks to some of the Director's Cuts. And there were a few other little scenes. I mean, nothing hugely dramatic, but still. Anyway, I've watched all of that, all those movies several times, and all the follow-ons. Anyway, so it is with some anticipation that tonight Lorrie and I will be watching the first two episodes of the Ridley Scott-produced, big-budget "Alien: Earth" series. |
| Leo: Oh. I didn't even know about this. |
| Steve: Which will be premiering, it'll be premiering on FX, FX on Hulu, and internationally on Disney+. So if you have access to FX or Hulu, the first two episodes are there. Now, the first one of these two was pre-released a few weeks ago during San Diego's Comic-Con, which garnered the series a remarkably high IMDB rating of 8.8. Which, you know, over the years I've found IMDB ratings to generally be useful, with 7.0 being my threshold for watchability. You know, if it's much below 7, it's like, eh, it may not be worth the time. But if it's 8.8, we may have another winner. |
| Leo: Okay. |
| Steve: The first season is eight episodes. So seven weeks from today, on September 23rd, the 8th and final episode of the first season will be released. Oh, and Leo, one of the stars is Timothy Oliphant, who I also really like. |
| Leo: What's the rough plot? |
| Steve: So what I don't understand is that it is two years before the Nostromo encounters the alien in space. So it's actually set two years before the events of the first film. But an alien-infected starship crash lands on Earth, thus "Alien: Earth." And we also have - we have a young egocentric greedy corporate techie who's trying to create a more perfect human or something, and so he wants the technology that the alien represents, of course. So anyway, it's meant to be many years. Apparently they spent more than $250 million on this. So, I mean, it is big-budget, lots of special effects. Trailers are online so anybody who's curious can go - maybe by the time you're hearing this, the whole, the first two episodes are already available. I mean, I guess they're available right now; right? Because it is Tuesday, August 12th. But anyway, I just wanted to let everybody know. I've always been a fan. Some of the follow-on movies were kind of marginal. |
| Leo: I liked "Prometheus." |
| Steve: I didn't understand... |
| Leo: There were so many. |
| Steve: I didn't understand "Prometheus." I still don't know what that was [crosstalk]. |
| Leo: What the hell happened there? |
| Steve: Yeah, I don't, you know, I guess I could read the book. And the most recent one was kind of good, "Romulus." |
| Leo: I liked "Romulus," yeah, yeah, yeah. |
| Steve: Yeah. So anyway, we have a big-budget, eight-episode, it's planned to be a multi-year series. So let's hope it's good. |
| Leo: "Alien: Earth." |
| Steve: "Alien: Earth." |
| Leo: Launches tonight. |
| Steve: Yeah. And I'm hoping that it's not really over the top horrific because I'll have a problem with Lorrie. You know, I mean, nobody wants to go through that right before bedtime. So she saw some of the preview. She saw the - she watched the trailers with me. And it does look really promising. |
| Leo: Well, it's by Noah Hawley, who did the "Fargo" TV series. |
| Steve: Yes. |
| Leo: Which is one of the best TV series ever. |
| Steve: I agree. And of course Timothy Oliphant... |
| Leo: [Crosstalk] has a great sense of the weird, too, which is nice, yeah. |
| Steve: Yeah. Timothy Oliphant, the actor, I really like. He of course played the Marshal on "Deadwood," and he was Raylan Givens on FX's "Justified" series. |
| Leo: Oh, he was good in that. All right. Yeah, I know who you're talking about. |
| Steve: Yeah. |
| Leo: Yeah, yeah. |
| Steve: Yeah. He plays a cyborg. |
| Leo: Oh, how funny. |
| Steve: Yeah. So... |
| Leo: Oh, I'm looking forward to this. Okay. |
| Steve: I am. Yeah, yeah. Okay. Last break, and then we're going to talk about Perplexity's Duplicity. |
| Leo: Perplexity's Duplicity. Not that easy to say, so I'm glad you said it, not me. Now, back to the man of the hour, Mr. Steve Gibson. |
| Steve: Okay. |
| Leo: And Security Now!. Soon as I get my mouse to respond. There we go. Go ahead. |
| Steve: So as I said, this is one of those news bits that began in today's show notes as just another piece of news up at the top. But the more I dug into it, the more I appreciated its significance. |
| Leo: Yeah. All right. I'm very interested in this because I'm a heavy Perplexity user. |
| Steve: As everyone will see, the deliberate and extensive behavior Cloudflare uncovered and discovered of Perplexity's Internet behavior is significant. And I should mention, I did not think to say this when I was going through this the first time, there's no reason to believe Perplexity is unique in this way. I mean, this is misbehavior. But we know OpenAI is not doing it. We'll talk about that. But we don't know who else might be. So it isn't my intention to single them out as the only bad guy around. Okay. But as we know, controversy emerged right alongside the appearance of the new capabilities of generative AI. The images being generated upon request often bore striking similarity to the known - and copyrighted - work of human artists. Poems and music sounded eerily familiar to those who were familiar with other original works of writers and musicians. So there was something familiar about this; right? And before long we began to realize that when massive large language models are trained on the Internet's content, all of which up to that point had been created solely by the application of human effort and creativity, anything that a generative AI might spit out was inherently a derivative work. Although it wasn't so directly, what we created carried a whiff of plagiarism. And in many cases it was much more than that. News sites began seeing the recognizable content of their human reporters appearing in the answers being offered by AI chatbots. None of this sat well with human creators, who wished to receive recognition and support in return for their life's work. The solution was to deploy the Internet's well-established automation controls to exclude these web scraping agents from websites that had no interest, thank you very much, in having their content absorbed by and used to train massive AI in the cloud. One of this TWiT Network's own sponsors, which catered to programmers of all ilk, advertised their AI bot-blocking as a feature, so that their users could feel confident that their collaborations would remain theirs and not be leaked out into the ether to become the unpaid-for property of this new generation of rapacious AI sponges. The Internet has a long history of bots. The bots most websites want and even actively invite and solicit are those belonging to search engines. Search was the early breakthrough application that entirely transformed the web. What good was it to create a quantity of terrific content of any kind if only your friends and family would ever be aware of its existence? Search changed all that. But the reason search bots were wanted is that search engines would list links back to the sites containing the desired content. So search bots would indirectly drive human traffic to the website where humans would see where that link led them, and then perhaps poke around and discover other goodies, all the while being presented with advertisements that were producing supporting revenue for the destination website. By comparison, AI model-building bots are not indexing a site for later discovery and linking. They're proactively scraping up all of the site's content, every juicy little morsel, and then feeding that original content into a massive AI model in the cloud. Effectively, the entire site's content is being incorporated into the AI model so that no one will ever need to visit that site again. They'll simply be able to ask the AI to obtain a homogenized and digested version of that site's once exclusive knowledge and wisdom. They say that imitation is the sincerest form of flattery. That may be true. But flattery doesn't pay the bills, nor does it give credit to an idea's originator. The consequence was that websites the world over quickly moved to closing their doors to content-sucking, model-building, AI scraping. Thanks, but no thanks. It quickly became clear that AI bots were the exact antithesis of search bots. Whereas search bots serve to drive future traffic back to a site, as I noted, the effect of an AI bot's visit is to reduce that site's future traffic. The problem was, for any AI modeler, the entire Internet was the biggest and juiciest source of free, ready-to-go, machine-readable content imaginable. You couldn't design a better source of training knowledge. Oh, sure, there was that annoyance that the Internet, being a product of humans, also contained a large amount of nonsense and crap and that the AI was no more able to tell the difference between truth and fiction than most of the humans who were consuming it. But none of that prevented the AI developers from turning their new scrapers loose on all that material to see what would happen. And then, of course, AI became totally and utterly dependent upon that big juicy flowing source of ever-changing knowledge. The fundamental problem of the entire approach is that it operates to model knowledge that it doesn't own and in the process does not again need. So AI inherently takes from the world's websites without ever giving anything back. So as those website doors began closing to AI, it had a problem. What it needed, what it had selfishly grown utterly dependent upon, was being denied. Increasingly, everywhere its bots turned they were encountering "bots.txt" files sitting passively in the roots of website domains that were permitting and welcoming the world's known search engines to enter, but were also denying entry to anything and everything else, and definitely to anything that hinted at being AI. Thanks anyway. Go suck out someone else's brain without even saying thanks. The problem is, the "robots.txt" file concept was created to help out bots. It's there for their own good. So it inherently depends upon the honor system. It's a file that, by convention, sits in a website's root directory. Shortly after I added the ShieldsUP! facility to GRC, I added a robots.txt to my site. GRC's site contains a bunch of automation that makes no sense to index. The DNS Spoofability pages, the ShieldsUP! service itself, our extensive Internet port references and much more just cause bots to become all tangled up and lost. So GRC's "robots.txt" page lets bots know beyond where dragons lie. Importantly, the "robots.txt" file is informational only. It is not any sort of enforcement mechanism. If some bot never looks at that file, or chooses to ignore its warnings, it might become tangled up in endless link-chain loops, pulling reams of nonsense data, and wasting a great deal of its own time and resources. But if so, so be it. It's been warned. As a consequence, search engines are thankful and appreciative of these "robots.txt" files. They figure that since their presence and services are a benefit to the site's management, anywhere a site doesn't want them to go is fine with them. But as things have evolved with AI, this is not the case. A visit from an AI scraper is not seen as offering the same benefit to a website as a visit from a search engine. Search engines are visiting to find and index. AI is here to steal. Last week, Cloudflare posted the news under the headline: "Perplexity is using stealth and undeclared crawlers to evade website no-crawl directives." Cloudflare wrote: "We are observing stealth crawling behavior from Perplexity, an AI-powered answer machine. Although Perplexity initially crawls from their declared user-agent, when they are presented with a network block, they appear to obscure their crawling identity in an attempt to circumvent the website's preferences. We see continued evidence that Perplexity is repeatedly modifying their user-agent and changing their source addresses to hide their crawling activity, as well as ignoring or sometimes failing to even fetch robots.txt files. "The Internet as we have known it is for the past three decades rapidly changing. But one thing remains constant: it is built on trust. There are clear preferences that crawlers should be transparent; serve a clear purpose; perform a specific activity; and, most importantly, follow website directives and preferences. Based on Perplexity's observed behavior, which is incompatible with those preferences, we have de-listed them as a verified bot and added heuristics to our managed rules that block this stealth crawling. "So what happened? We received complaints from customers who had both disallowed Perplexity crawling activity in their robots.txt files and also created WAF - Web Application Firewall - rules to specifically block both of Perplexity's declared crawlers, PerplexityBot and Perplexity-User. These customers subsequently informed us that Perplexity was still able to access their content even when they saw its bots successfully blocked. We confirmed that Perplexity's crawlers were in fact being blocked on the specific pages in question. We then performed several targeted tests to confirm what exact behavior we could observe." Get this, Leo. "We created multiple brand-new domains, similar to testexample.com and secretexample.com. These domains were newly purchased and had not yet been indexed by any search engine nor made publicly accessible in any discoverable way. We implemented a robots.txt file with directives to stop any respectful bots from accessing any part of a website. We conducted an experiment by querying Perplexity AI with questions about these domains, and discovered Perplexity was still providing detailed information regarding the exact content hosted on..." |
| Leo: Wait a minute. Did they ask by name for those domains? Did they specifically say to Perplexity, tell me what's on those domains? |
| Steve: I don't know. |
| Leo: Yes, they did. |
| Steve: Okay. |
| Leo: So everything they just said about, well, we made a new thing, and it never could have known, they told it where to go. |
| Steve: They also blocked it in the robots.txt file. |
| Leo: But that's not what robots.txt is for. It's to keep your domain or pages from your domain out of the search index. This is a very controversial subject, Steve, and I think you might be misunderstanding. I think Cloudflare is wrong on this one. But I'll let you finish, and we can talk about it at the end. But I think it's more complicated, it's more nuanced than that. If you create a site, and then ask an AI to go look at that site, just as you would ask a browser to look at that site, you can't complain when the AI goes to look at the site. If you do, you're going to break all agentic AI, all MCPs, all research AI, same thing. |
| Steve: So one problem, you're saying the problem is that they created a site, and they asked Perplexity. |
| Leo: Yes. |
| Steve: But that's not the abuse case. They created the site, and anybody could ask Perplexity. |
| Leo: Right. If you ask a browser to go to a site, and the browser shows you the site, that's how it's designed. Perplexity is very clear. They said we are not looking at the site to train on it. We're not training on that site. But the user asked us for a summary of that site, so we gave them that. It's like you going there on a browser. That's not what they were doing. They weren't training on it. I understand if you say, oh, they're just training on every site they can find. By the way, if you didn't have AI train every site that you could find, it wouldn't be very useful, either. |
| Steve: So unfortunately, this doesn't say either way, whether they were training or not. |
| Leo: No, Perplexity had a response, which you should read. |
| Steve: Okay. |
| Leo: Because they say they were not. |
| Steve: Oh, okay. Good. Well, in that case, that's significant, although Perplexity goes to great lengths to avoid - okay. So let me just finish explaining... |
| Leo: Yeah, finish, yeah, yeah. Yeah, yeah. |
| Steve: ...what Cloudflare saw. So they said: "We conducted an experiment by querying Perplexity AI with questions about these domains, and discovered Perplexity was still providing detailed information regarding the exact content hosted on each of these restricted domains. The response was unexpected, as we had taken all necessary precautions to prevent this data from being retrievable by their crawlers." |
| Leo: Except for removing it from DNS. If they hadn't published the domain, Perplexity wouldn't have found it. They couldn't have asked for it. |
| Steve: Well, it's a valid website, Leo. |
| Leo: Yes. |
| Steve: So just hold on. Let me finish. |
| Leo: It's a valid - it's very important, though. It's a valid website. |
| Steve: Wait till you hear - okay. They said: "Our multiple test domains explicitly prohibited all automated access by specifying in robots.txt and had specific Web Application Firewall rules that blocked crawling from Perplexity's public crawlers." Their crawlers. "We observed that Perplexity uses not only their declared user-agent" - fine - "but also a generic browser intended to impersonate Google Chrome on macOS when their declared crawler was blocked." So they are explicitly and deliberately bypassing someone's attempt to block them. We have two user-agent strings in the show notes, which Cloudflare shared, the standard Perplexity.ai user-agent and then the stealth one, which pretends to be AppleWebKit Safari. They said: "Both their declared and undeclared crawlers were attempting to access the content for scraping, contrary to the web crawling norms as outlined in RFC 9309. This undeclared crawler utilized multiple IPs not listed in Perplexity's official IP range, and would rotate through these IPs in response to the restrictive robots.txt policy and block from Cloudflare. In addition to rotating IPs, we observed requests coming from different ASNs in attempts to further evade website blocks. This activity was observed across tens of thousands of domains and millions of requests per day. We were able to fingerprint this crawler using a combination of machine learning and network signals." Okay. So Perplexity, Leo, is deliberately and purposefully working hard to bypass a website's web management's own decision whether they want Perplexity to have anything to do with their site or not. And I would argue that any website has that right. |
| Leo: It raises - look, I'm not going to say I know for sure what the answer is because there are all sorts of problems with crawlers bringing sites down. Wikipedia's had a huge problem with crawlers doing that. In my mind there's a difference between training on a site and going to a site because a user requested information about that site from the AI. Anytime you use a research AI, if you go and say, hey, tell me more about Vitamin D to ChatGPT research or DeepSeek or any of these, if they're using the modern web, they're basically doing the same thing, which is they're going to a bunch of sites. They're going to give you, you know, a source on the site and summarize it. Now, we can decide that we don't want that to happen. I think you need more than robots.txt. But let's say you have - you use robots.txt, and you're saying you don't want them to do that, that's going to very much impede agentic and MCP usage. It's going to impede research AIs. It's going to impede all of... |
| Steve: Who cares? It's the site's right. Cloudflare is saying that their customers said we don't want Perplexity on our site. And so they put rules in, following Perplexity's guidelines. Perplexity lists the IP ranges that they come in from. Perplexity has a user-agent... |
| Leo: And they went around all of that. |
| Steve: Yes. And Perplexity did not abide by the block. And not only that, they went to extreme - they managed to generate IPs from other top-level ISPs using different ASNs because even using a different IP within the same top-level provider could be seen as a signal. So they've gone out of their way to evade the clear request by a webmaster, "You are not welcome here." And I think that's the end of it, period. Anyone has a right to say I don't want you visiting my site. |
| Leo: Does anybody have a right to say, "If you don't see my ads, I don't want you to see my site?" I guess they do; right? You could say you've got to have the ads, or you can't see my site. |
| Steve: Many people do that, yeah. |
| Leo: An ad blocker does that. An ad blocker's deciding what part of the site to display and what part of the site not to display. Violating probably in many cases the desire of the site's owner. And yet we use ad blockers. I think there's an issue here because the real question is, is an AI company different than a browser? If I decide to use an AI to look at a website versus a browser to look at a website, you're saying the webmaster should be able to distinguish between the two. And I say yes, that's probably true. But it's not as clear-cut as you say because I'm just using a different tool, a browser-like tool, to look at that site. Maybe I want to aggregate information about Vitamin D from 20 sites. I could manually go to each site, or I could ask a tool to go out and look at those sites and aggregate that data. |
| Steve: So certainly, if the site were to block access by Perplexity, then Perplexity would not be, I mean... |
| Leo: Couldn't include that site in its sources. That's right. |
| Steve: Exactly. And so that would be the site's loss, if they wanted to be included in Perplexity's sources. Apparently, some of Cloudflare's customers don't want to be, I mean, I'm not here to psychoanalyze them. I don't know what they're thinking or why they don't want to be part of that. It may be because they feel that AI is not in their benefit. I mean, we had that advertiser who was boasting about the fact that they keep AI off the site. |
| Leo: Right. |
| Steve: They didn't want AI, you know, sucking in all of their customers' dialogs. |
| Leo: Again, there's a difference between training and going there to read the site, though. Yes? And I think people conflate the two. I understand you may say, if you're Reddit or The New York Times, I don't want AIs to train on my content. But I think it's a very fine line to say I don't want a person to go read my content versus an AI. And this is a complex legal issue. It goes back to the First Amendment and the right to read. We talk about this a lot on IM, which is probably where we should move this conversation. But Cathy Gellis, who's a constitutional lawyer, has often talked about, if you start restricting the right of AIs to read, it undermines a fundamental First Amendment right, the right to read. And that's problematic. |
| Steve: Wow. |
| Leo: Yeah. So remember, we're not training here. We're asking for the contents of the site on behalf of the user. Now, Cloudflare says it did everything to prevent that, except it published the site publicly. |
| Steve: Well, everyone has a right to publish the site publicly. |
| Leo: Yeah. So don't I have the right to use the browser of choice? And if my browser happens to be Perplexity as opposed to Firefox? |
| Steve: I guess my only problem is that very clear rules were put up, which Perplexity themselves said they would honor. |
| Leo: Right. |
| Steve: The site said we don't want you, so we're blocking you. And then Perplexity demonstrated that they will go to extreme measures to circumvent that preference stated by the site. |
| Leo: Right. |
| Steve: So, you know. And, for example, OpenAI doesn't do that at all. They ran the same test against other AIs that honored all of the rules. |
| Leo: That's not true, by the way. Because OpenAI, when it does research, does do that. And I suspect that Cloudflare, to be honest, was more interested in getting attention to this than they were in actually doing this deep research. And that's certainly the claim that Perplexity makes. I think it's complicated. I don't know what the right answer is. You're right. One of the things Perplexity said is that Cloudflare's diagram of how Perplexity works was incorrect, was just fundamentally wrong, and they never asked us. That's not how it works. Perplexity said Cloudflare's either dangerously misinformed on the basics of AI, or simply more flare than cloud. Remember that Cloudflare has a vested interest in people saying, you know, using its tool. |
| Steve: Absolutely. I get it. Yup. |
| Leo: Even more embarrassing, Perplexity said, Cloudflare published a technical diagram supposedly showing Perplexity's crawling workflow that bears no resemblance to how Perplexity actually works. If they were interested in understanding the data, they could have done what we encourage all Perplexity users to do: just ask. |
| Steve: Okay. So that's not fair. I saw the diagram. I studied the diagram. I understand the diagram. The diagram is correct. The diagram is something different than what you're reading that Perplexity is saying. |
| Leo: Okay. |
| Steve: What Cloudflare showed in that diagram was the exact behavior they observed. So that is a diagram of observed behavior based on when Perplexity's bot was blocked and what they then saw Perplexity do as a consequence of the blocking. |
| Leo: Ah. So that's because Perplexity occasionally uses something called "Browserbase," which is an automated browsing service. They say: "Cloudflare misattributed 3-6 million daily requests from Browserbase to us." |
| Steve: That's entirely reasonable, yes. |
| Leo: Because we don't use all - we didn't use and do not use all of Browserbase's capacity. We use a small fraction of it on occasion. |
| Steve: And that could also account for the differing ASN sources of the IPs. |
| Leo: Exactly. It's just coming from Browserbase. |
| Steve: In Browserbase it's coming from a different top tier provider. |
| Leo: I don't know what the answer is. I read the Cloudflare thing, and I was as incensed as you were. And then I read Perplexity's response, and I thought, well, this maybe is more complicated. But there is also this fundamental issue of... |
| Steve: Yeah, that's really interesting. It's interesting, I mean, that the First Amendment protects AI... |
| Leo: The right to read. Well, it protects - see, the courts haven't said. It protects our right to read. And Cathy Gellis's concern is, if the courts say, well, AIs don't have the right to read, that it may well impinge on our right to read. |
| Steve: Has it been tested? |
| Leo: Not yet. I think that what's important is there is this concept of the open web, the idea that you put something on the web, you're putting it out there for everybody to read. Now, I understand there's a lot of issues around this - bandwidth issues. |
| Steve: All the economics that have developed. |
| Leo: Economics, ads and all of that. So this has really complicated it. But fundamentally, I'm a believer in the open web. The whole point of the web is putting that information out there for everyone. And I think it's discriminatory to say, well, there's only one way you can look at my website, and that's in Firefox or Chrome. What if some website said we don't want anybody with Firefox to visit our website? They'd have the right to do that, but it's just fundamentally... |
| Steve: Yeah, I think... |
| Leo: But it fundamentally to me undermines the concept of an open web. |
| Steve: Yeah, but then, you know, what about a paywall? |
| Leo: Right. What we're really talking about is property rights. And I think the web - I'm a hippie. |
| Steve: Did you see that Perplexity has offered to buy Chrome? |
| Leo: Yeah, that's just a publicity stunt because actually they're offering more money than they're worth. |
| Steve: Yes, 34 million, or billion, rather, yeah, yeah. |
| Leo: Yeah. You know, it's - Perplexity, what Perplexity does it does better than anybody else. I've used Perplexity for two years, and I really like it. I have a paid account with Perplexity, but of course I have a paid account with every single AI account I could get, including Kagi. But Perplexity, Kagi have assistants that do this kind of aggregating. And I think it's incredibly valuable. And I think we have to ask ourselves whether we want the open web to be truly open, or whether we want it to be fiefdoms that are controlled by property owners. |
| Steve: So does Perplexity - Perplexity does no training? They're not an AI, just to have the model? |
| Leo: Well, they use other people's models for the most part. They are a front end to a variety of models. In fact, when you're using Perplexity you can use a variety of different models, including ChatGPT. |
| Steve: Okay. But are they the trainers of those models, I guess is the question. |
| Leo: No, no, no. They're tuning them, maybe. I don't know. But no, they're primarily - they started as being an aggregator of other people's models. |
| Steve: So like calling into, like, to OpenAI's API and getting the results of its training? Or are they training themselves? |
| Leo: Yes. The former. |
| Steve: Huh. |
| Leo: Now, I think over time Perplexity has added some additional LLMs of its own. And also I don't - they're not completely forthright on exactly how they're combining AIs and so forth. But the whole idea of Perplexity was it was a meta AI. It was an AI that let you choose the model - Anthropic or OpenAI. |
| Steve: Well, and it would be interesting to know why these Cloudflare customers are not wanting Perplexity to visit their site. |
| Leo: Well, remember, this wasn't Cloudflare customers. This is Cloudflare setting up dummy sites. |
| Steve: No, no, no. This was instigated by Cloudflare customers... |
| Leo: People complaining? |
| Steve: ...who put up all the blocks themselves and then Perplexity got around them. |
| Leo: Right. |
| Steve: And the they complained Perplexity is not honoring our request for it not to save our site. |
| Leo: Okay. So this is in - this was in response to... |
| Steve: Yes, and then Cloudflare investigated. |
| Leo: Yeah. And maybe investigated making some mistakes. I'm not sure. I think it's very complicated. And I think we really have to think about what our fundamental values are when it comes to the web, how open we want the web to be. Or do we want to carve it up into little fiefdoms? And if we do, then we should probably abandon the concept of research AIs, and AIs in general. Because if AIs can only train on a very limited set of information, they're going to be less [crosstalk] value. |
| Steve: Well, and it's very clear that this does upset the traditional economics. |
| Leo: Well, so does Google. I mean, we're at Google Zero; right? Google no longer sends site traffic. Well, they claim they do, but I don't believe Google sends as much traffic as they used to. |
| Steve: Right. |
| Leo: And by the way, so do paywalls. I mean, they also cut your site traffic. So it gets - it's complicated I guess is my answer. And I don't fully credit what Cloudflare says. I think it's more complicated, more nuanced. That's all I wanted to say. |
| Steve: Okay. |
| Leo: I think you make an excellent point, that ultimately a property owner should be able to decide who enters their property. But then we have to ask the question, well, is that what we want on the web is a bunch of private property? |
| Steve: That's what we have. |
| Leo: It is what we have. |
| Steve: That is what has happened. You know, and now we're putting age restrictions and laws behind it. |
| Leo: Yeah, none of that seems to me to be the right direction. |
| Steve: Yeah. |
| Leo: I feel like we could have more value out of the web if it were more open. |
| Steve: Just, you know, the Internet's become a microcosm of our world. |
| Leo: But this is why we do podcasting in the old-fashioned way. This is why we're not exclusive on Spotify or Audible or Amazon. But we do RSS feeds because it's open. |
| Steve: Yup. |
| Leo: Right? And I'm a, see, I'm a believer in the open web. We do creative commons content. My blog is open. I don't have any robots.txt on it because I think that's the right way to go. But I understand not everybody feels that way. And we need to figure out there are competing needs in this. |
| Steve: That's what makes all this interesting, my friend. |
| Leo: It is. It's what makes it interesting, isn't it. Anyway, thank you for letting me interrupt. I appreciate it. |
| Steve: No, no, I'm done. I think we've had a great conversation about this. And I'm glad you brought that view because it was, you're right, mine was single-sided and Cloudflare-centric. |
| Leo: Yeah. Read what Perplexity says. There's a lot of debate on the 'Net. We've talked a lot, we talked last week extensively on Intelligent Machines about it. And I don't know what the right answer is. I do agree with you. A property owner should be able to build a fence that nobody can go over. I agree with you. I don't think it's a way to divide the web up, but if that's what happens... |
| Steve: Yeah. And I guess I'm a bit of Cloudflare fanboy. I really do believe those guys. |
| Leo: I agree. |
| Steve: Everything they do is, you know, they know their technology. They know the web. I just think they're on the right side. And so, you know, that may have biased me toward them. But, you know, Perplexity has something to lose, too. It's needing to respond to this. I just don't know, you know, how open they've been. |
| Leo: Right. Perplexity says when companies like Cloudflare mischaracterize user-driven AI assistance as malicious bots, they're arguing that any automated tools serving users should be suspect, a position that would criminalize email clients and web browsers or any other service a would-be gatekeeper decided they don't like. And then you're saying gatekeepers should be able to keep their gates. |
| Steve: Yes. And they specifically list the range of IPs that their bots operate from. |
| Leo: Right. |
| Steve: And so if someone says thank you for that... |
| Leo: Oh, that's - but this is that third-party tool that they were using, and that's part of the problem, yeah. |
| Steve: Yeah. |
| Leo: And I think that's where the real disconnect was, that Cloudflare misunderstood Browserbase and how it was... |
| Steve: I would be interested in Cloudflare's rebuttal to Perplexity's rebuttal because, I mean, I'm just - I'm academically interested. But it does seem that, you know, when Perplexity's named user-agent is denied, suddenly a non-Perplexity... |
| Leo: Right. |
| Steve: Safari, you know, Safari... |
| Leo: There is a history, too, of Perplexity doing - this is the Wired story from June 19th. |
| Steve: Not mincing any words. |
| Leo: Not mincing any words. There is a history of Perplexity doing some shady things. So I'm not saying Perplexity is absolutely in the right on this. I'm saying it's a larger subject than just Perplexity versus Cloudflare. It's about what we want the web to be, kind of. Anyway, great show. |
| Steve: We've got the world we want. |
| Leo: Yeah, well, I don't know if we get the world we want. We get the world we get. |
| Steve: Well, the world we deserve, maybe. |
| Leo: Maybe the world we deserve. Steve Gibson is at GRC.com. Interestingly, you don't block anything from your site, do you? |
| Steve: No. No. |
| Leo: No. Do you have a robots.txt? |
| Steve: Only to keep ShieldsUP! and DNS Spoofability because you don't want Google trying to index something that is checking a user's bandwidth connections. |
| Leo: Right. Steve, that's a very good use for robots.txt and should be absolutely honored. |
|
 | Gibson Research Corporation is owned and operated by Steve Gibson. The contents of this page are Copyright (c) 2025 Gibson Research Corporation. SpinRite, ShieldsUP, NanoProbe, and any other indicated trademarks are registered trademarks of Gibson Research Corporation, Laguna Hills, CA, USA. GRC's web and customer privacy policy. |  |
| Last Edit: Aug 19, 2025 at 09:00 (117.78 days ago) | Viewed 8 times per day |