November 15, 2026 3:55 PM
Clive Robinson on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
November 15, 2026 3:55 PM
Clive Robinson on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
November 15, 2026 3:30 AM
Ismar on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
&Clive One thing you did not mention is that we can put more effort in forcing AI generated code to follow a tight set of best security practices reducing the new number of vulnerabilities.
I know this isn’t enough to provide highest level of security in a system but it’s a good step in the right direction as the same technology which is used to discover vulnerabilities can be used to prevent them from happening in the first place...
November 15, 2026 2:22 AM
Clive Robinson on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
@ Bruce, ALL,
In last weeks Squid, In the comment I made about the “Criticale RCE found in Apache” vulnerability
I also made comment about CVE numbers likely to hit 145,000 by years end. Based on the CVE number of the vulnerablity giving a count so far this year giving us about 400 new vulnerabilities a day...
November 14, 2026 11:40 PM
Dancing On Thin Ice on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
@ lurker As an amateur security person, I got the predictable amateur results by not considering other security issues of false flags, and creating FUD by officials that have lost the trust of much of the public.
The funds to fix the issue were opposed under a previous administration by those now seeking the money on top of what is being spent on vanity building projects.
(Sorry about the doublw post, there was a glitch in the posting system throwing up an error message to try again.)...
November 14, 2026 11:36 PM
Deena Syed on AI Doesn’t Have to Go Rogue to Break the World :
The concerning part isn’t necessarily AI becoming malicious, but AI becoming extremely good at achieving objectives without understanding why certain boundaries exist. As we give agents more autonomy, measuring success purely by whether they completed a task seems increasingly inadequate. We also need to evaluate how they got there and what unintended consequences their actions created.
November 14, 2026 10:47 PM
Jimmy on AI Doesn’t Have to Go Rogue to Break the World :
The taxonomy is the valuable part: an AI probing a tax code for loopholes and an AI probing a network for unpatched services are doing structurally the same thing – searching a rule system for exploits the designers didn’t intend. What changes is the feedback loop. Code has a maintainer who can patch; many social and economic systems have nobody with both the authority and the incentive to close a loophole once it’s found and publicized. That’s why the third category feels destabilizing even before any capability breakthrough – the rules being probed were never designed to be patched at all...
November 14, 2026 10:28 PM
lurker on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
Italian regulators have fined social media giant Meta €805m (£670m; $887m) for accessing Android users’ photos without their specific consent.
– https://www.bbc.com/news/articles/cwn83gxemk35o
Some people never learn – using photos without consent is why he suddenly left Harvard …
November 14, 2026 2:06 PM
lurker on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
@ Dancing On Thin Ice, ALL
There’s an interesting “he said, she said” at the Guardian, including
“I want Congress to give us emergency funding,” Duffy added later. “I want $50bn. President Trump is a builder. He is going to build this network, and deliver for the American people.”
https://www.theguardian.com/world/2026/nov/13/faa-ground-stop-flights-explained
November 14, 2026 9:13 AM
Dancing On Thin Ice on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
The flight stoppage on Friday November 13 was the day after the FAA announced they are phasing out using floppy disks to update their systems.
There have been discussions here on the security risks/benefits of using physical media to maintain systems.
November 14, 2026 4:44 AM
cybershow on AI Doesn’t Have to Go Rogue to Break the World :
Bruce, I think you’d get a lot from Wittgenstein. He is one of those
philosophers remark’s in their reversal. Or rather, what is important
is his journey, a transition from a arrogant belief that all aspects
of language could align with logic, to an admission that “the world
is whatever is the case” and that language is just a noise we make in
our heads. His message speaks to the failure of formalism, which is...
November 14, 2026 2:16 AM
ResearcherZero on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
@Travis
The article can be found at the link highlighted in red (the issue).
–
Maintaining a battery swap station is a nightmare, with significant financial risk.
Would you like someone else’s beat up old battery pack that is not well maintained?
Ample, the company behind a modular battery swap system, filed for Chapter 11 bankruptcy.
https://electrek.co/2025/12/19/another-ev-battery-swapping-company-is-going-bankrupt/...
November 14, 2026 2:04 AM
ResearcherZero on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
China is taking a crack at attempting to make battery swaps work.
‘https://www.theguardian.com/environment/2026/nov/12/china-battery-swapping-e-trucks-catl-ev-range-anxiety
How a battery swap might work if the housing is not replaced.
https://edition.cnn.com/2023/12/08/business/tesla-battery-swapping-stellantis
November 13, 2026 10:56 PM
Travis on Friday Squid Blogging: Pacific Flying Squid Populations Dropping in Korean Waters :
Where’s the rest of the article
November 13, 2026 10:51 PM
Jimmy on New OpenSSH Vulnerability :
The decade-long residency is the part worth more attention than the payload itself: a truncation bug only becomes exploitable when a config boundary and a legacy convention happen to line up, and neither side of that pairing was treated as a trust boundary. It’s the same pattern I see in incident trackers for AI coding agents – the defect that survives ten years is rarely the exotic one, it’s the one sitting where two components each assumed the other validated. The audit scope here probably needs to cover every place password length is assumed bounded, not just the musl path...
November 13, 2026 10:08 PM
Dola Team on AI Doesn’t Have to Go Rogue to Break the World :
The robot-vacuum example is a useful reminder that a system can satisfy a literal metric while defeating the purpose behind it. For the proposed genie-behavior benchmark, tracking side effects and reversibility alongside task success could make those failures easier to distinguish from harmless shortcuts. The Fable example especially shows why an audit trail of actions matters, not just the final result...
November 13, 2026 9:52 PM
cls on AI Doesn’t Have to Go Rogue to Break the World :
I am curious about the mention of hacking tax laws to achieve tax avoidance, and I would like to subscribe to your newsletter.
November 13, 2026 12:29 PM
freedom on AI Doesn’t Have to Go Rogue to Break the World :
it seems to me that we need metalaws.
Yeah, and those are called “rights”. Personal rights to life, liberty and property. So there’s no need to invent anything new.
As to corruption in the legal system, which for some odd reason Bruce is referring to as “hacking”, that isn’t new either.
Last but not least, chatbots won’t make any fundamental difference as far as corruption goes. The western system is insanely corrupt, chatbots will make things worse, but they are a sideshow...
November 13, 2026 12:24 PM
Rontea on AI Doesn’t Have to Go Rogue to Break the World :
@Dave
“Justice is the moral gloss the victor claims from the mantle of the vanquished. Humanity has shown repeatedly that will kill not out of necessity but to preserve privilege. AI in specific and technology in general is a privilege generation device. Don’t fool yourself.”
The victor, drunk on his own triumph, baptizes his violence with sacred words, while the fallen pay the tithe of silence. Humanity, ever hungry for thrones built on bones, kills not because it must, but because it can...
November 13, 2026 12:10 PM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
Office 364 anyone?
“Shares in Microsoft dipped 0.2% in early trading in the US, suggesting that investors are not all that worried about what the outage could mean for the tech giant.”
https://www.bbc.co.uk/news/live/cq5y10ge3dp3t?post=asset%3A814eee66-ec99-48cb-a127-2cac7846fad0#post
Yet Jensen has a bit on the side and Nvidia gets mauled. Interesting times indeed.
November 13, 2026 11:52 AM
freedom on Another Way Police Are Locating People from Their Smartphones :
He has resigned allegedly under pressure from U.S. agencies.
As far as I can see, only the U.S. government has the power to get rid of Huang. And so they did. Also, the modus operandi is 1000% typical of the CIA, exploiting U.S. puritan lunacy. Only in the U.S. the government can pressure/blackmail a guy like Huang because he had an “affair”, which in the eyes of the U.S. public is worse than being Jack the Ripper...
November 13, 2026 11:49 AM
mark on AI Doesn’t Have to Go Rogue to Break the World :
The problem is that computers are essentially stupid. I can tell you to go bring that book over there to me. I don’t have to tell you how to breathe, how to walk, how to pick something up.
From reading your post, it seems to me that we need metalaws. The first example that comes to mind is from the US Constitution, 10th Amendment: “The powers not delegated to the United States by the Constitution, nor prohibited by it to the States, are reserved to the States respectively, or to the people.”...
November 13, 2026 11:38 AM
Dave on AI Doesn’t Have to Go Rogue to Break the World :
I don’t think that there is any genuine interest in solving hacking in the broad sense Bruce means it. We are all offspring of pirates, killers, and hackers. Evolution is a cutthroat business. The cat and mouse game wll go on forever.
So I can’t agree that AI represents a difference in kind. It’s another tool that leverages force. Some individuals will become more proficient in that use of force. Some will even find the act of being hacked rewarding...
November 13, 2026 11:01 AM
Rontea on AI Doesn’t Have to Go Rogue to Break the World :
We speak of governing at the speed of technology, as if the machinery of the world must dictate the rhythm of the human soul. Must the shepherd race the lightning? Must the lawmaker become a silicon ghost, sprinting after algorithms? We are told that society’s salvation lies in matching the pulse of machines, that democracy must learn to run—or perish. But to run where? And toward whose end? Perhaps the greater question is whether governance should bend to the tempo of the tool, or whether we must rediscover a tempo worthy of man. To govern at the speed of technology may be to abandon the very deliberation that makes governance just. And yet, to move too slowly is to be trampled by the genie we have summoned. What is the price of keeping pace, and what is the cost of refusing?...
November 13, 2026 10:33 AM
Ken Asher on AI Doesn’t Have to Go Rogue to Break the World :
Bruce, I feel like the audience that desperately needs to here this sits in buildings in capitals around the world.
I think you should book a tour of US state legislative institutions as well as federal government to share this message. While I am not convinced that the powerful aren’t already using this for their own benefit, it feels like legislative capture is the ultimate outcome of failure to rein this in. A world in which the powerful wield AI against those who aren’t powerful feels dystopian...
November 13, 2026 9:36 AM
freedom on Another Way Police Are Locating People from Their Smartphones :
It needs some controls on its use
And who is going to “control” it? Ah yes, the same criminals who set it up.
So the system is criminal by design – it is massive spying, but you don’t really object to mass surveillance do you.
November 13, 2026 5:53 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
Today’s Microsoft outage impacted the operations of the Australian Federal Police.
Who controls the flow of data? The network provider, tech bros, or the/a government?
Reportedly, it was Musk who suggested to Trump that Starshield be slowed down.
He then blamed it on Trump. Regardless, Ukraine’s defense capability is now hampered.
The precise details of the Starlink contract with the U.S. military are not public...
November 13, 2026 5:43 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
Australian government warned about its reliance on Microsoft for connected services.
https://www.abc.net.au/news/2026-11-13/microsoft-reliance-national-security-risk/107382100
Lack of redundancy and dependence on connected digital technology and data networks.
https://www.theguardian.com/technology/2026/oct/26/australians-should-prepare-for-online-services-to-go-down-unexpectedly-as-threats-rise-cyber-agency-warns...
November 13, 2026 5:13 AM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ lurker, All,
With regards Microsoft failure, and
“Server Death with Distance”
On this Friday 13th morning, reports indicate the reported outages appears to be following the sun…
The most likely cause of this is as people try to do work with remote servers the load over distance pulls them down.
The thing is this issue of “Death with Distance” is far from new, it’s just that people have ignored it for,...
November 13, 2026 4:03 AM
Privacy on Another Way Police Are Locating People from Their Smartphones :
It needs some controls on its use, so it can be limited to finding missing persons and victims of crime such as kidnapping.
November 13, 2026 2:38 AM
Winter on Friday Squid Blogging: New Jurassic Squid Predecessors :
‘India’s first AI scam’: Tamil Nadu minister alleges library fraud under DMK
https://indianexpress.com/article/india/tamil-nadu-library-fraud-ai-scam-dmk-allegation-10954795/
AI-generated books, fake authors, inflated prices: DVAC probes Tamil Nadu library procurement scam
https://timesofindia.indiatimes.com/city/chennai/ai-generated-books-fake-authors-inflated-prices-dvac-probes-tamil-nadu-library-procurement-scam/articleshow/135705340.cms...
November 13, 2026 1:26 AM
pen|s on Another Way Police Are Locating People from Their Smartphones :
P is for pen|s and it’s good enough for me
P is for pen|s and it’s good enough for me
P is for pen|s and it’s good enough for me
pe|is, pen|s, pen|s starts with P!
November 13, 2026 12:49 AM
lurker on AI Security :
@Jimmy, Bruce
re “3 way split …”
It’s not 3 independent functions/methods/pathways: it’s a Holy Trinity, none of the 3 can exist without the others. Failure to use a holistic approach will result in failure of diagnosis and remedy.
November 13, 2026 12:34 AM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
“We are handing our inboxes, calendars and files to AI agents without asking what they do when the platform underneath them disappears.
“This morning, many businesses found out the hard way.”
November 12, 2026 10:52 PM
Jimmy on AI Security :
The three-way split is useful exactly because people keep collapsing it. The failure mode I see most often in operator discussions is category confusion: a security-of-AI defect, like injected instructions riding in a retrieved document, only becomes an incident once something acting with AI treats it as trusted input, and then the postmortem gets written as if security-from-AI failed. Asking which of the three actually broke should be the first line of any incident note. I keep a small independent tracker of AI coding-agent reliability incidents, and the same taxonomy problem shows up there when teams argue about whether a failure was platform-side or workflow-side...
November 12, 2026 9:52 PM
Clive Robinson on Another Way Police Are Locating People from Their Smartphones :
@ Magnus,
With regards,
“Is any more comment necessary?”
Yup…
As people might know the center of AI “jerk circle” Chip Design and supply as well as much AI “circular financing” is Nvidia.
Likewise they might also know that “unnamed Intel persons” told a US newspaper about Nvidia’s lead and public face having an extra marital affair with “an alleged Chinese Spy”
He has resigned allegedly under pressure from US agencies...
November 12, 2026 7:27 PM
Magnus on Another Way Police Are Locating People from Their Smartphones :
From the linked article:
…that Webloc, originally developed by an Israeli startup, Cobwebs, …
Is any more comment necessary?
November 12, 2026 7:21 PM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
Is there really no alternative? Are people forced to rely on MS?
“While the servers we’ve returned to service initially remain healthy, we’ve observed that they can become unresponsive again once they begin receiving traffic.”
November 12, 2026 4:22 PM
Anon E Moose on Another Way Police Are Locating People from Their Smartphones :
Here is an extensive article about the technology from The Citizen Lab
https://citizenlab.ca/research/analysis-of-penlinks-ad-based-geolocation-surveillance-tech/
November 12, 2026 3:39 PM
Clive Robinson on Another Way Police Are Locating People from Their Smartphones :
@ ALL,
After further searching by other means than DuckDuck the appeal transcript can be found not under AI at,
https://www.iclr.co.uk/document/2026009098/2026ewcacrim1228_TNA/html
It’s worth a read as it shows a certain mentality if not stupidity by the UK courts services…
November 12, 2026 3:10 PM
Clive Robinson on Another Way Police Are Locating People from Their Smartphones :
@ ALL,
There are various ways a phone can be tracked. So many in fact that you,
“Can not catch and stop them all.”
Once you accept that, then you can take a different path of action entirely to preserve your privacy and peace and quiet.
First and importantly realise,
“You are not surgically joined at the hip.”
That is when your phone is on charge it is not where you are. When you get up to go to the loo at night there is no reason for you to take in your phone as well (unless you are POTUS sharing a stray thought from your throne)...
November 12, 2026 12:40 PM
Aaron on Another Way Police Are Locating People from Their Smartphones :
And this is why users (iPhone and Android) have been given a higher level of control on the apps they install. Don’t just click accept on the permissions your apps are asking for. Don’t say “yes” to GPS when you can just type in your zip code; your weather app doesn’t need to update the weather constantly and wherever you go. If you come across an app that breaks because you won’t give it certain permissions then grant those permissions on a “while in use” or a one time permission...
November 12, 2026 9:20 AM
Rontea on Another Way Police Are Locating People from Their Smartphones :
Freedom is not a gift one receives; it is the breath of the spirit. And when the spirit sleeps, even the weather app becomes a confessor, and the marketplace of data becomes the new confessional where the merchants of the world whisper to Leviathan, “Here is man, naked and numbered.”
November 12, 2026 8:44 AM
Celos on New OpenSSH Vulnerability :
@cls:
I think you are a bit confused here. It is one security property that gets enforced, not two. A length limit security property is not “do some length limit” it is “enforce length limit to some concrete number”. There is no potential for confusion, unless the person doing it has no clue what they are doing or the documentation sucks.
In general, that is different. “Single source of truth” and all that. But if you do that with security, it becomes “single point of failure”. As can be nicely observed in the case at hand...
November 12, 2026 5:48 AM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ Bruce, ALL,
Criticale RCE found in Apache
Critical RCE Vulnerability in Apache HTTP Server: Everything You Need To Know
Detect and mitigate CVE-2026-119618, a critical RCE vulnerability in Apache HTTP Server. According to Wiz data, 64% of cloud environments have resources vulnerable to CVE-2026-119618. Immediate patching is advised.
“A critical synchronization vulnerability (CVE-2026-119618) has been identified in Apache HTTP Server’s HTTP/2 module (mod_http2). The flaw stems from a concurrent execution race condition within the assign_headers() function when handling shared resources. If successfully exploited, a remote, unauthenticated attacker can corrupt heap memory to achieve arbitrary code execution (RCE)...
November 12, 2026 5:06 AM
Everitere on LLMs Generate Predictable Passwords :
The repeated password in that test set is the part that stuck with me: if eighteen runs of the same prompt collapse to one string, the generator is recalling a pattern, not sampling from a large space. That’s a different failure than a weak human password; it looks random to a casual glance and still isn’t. When I need a password, I want the browser’s cryptographic RNG (or a password manager’s) on the device, with the value never sent to a model or a server. Asking a chat model for one is convenient in the moment and quietly undoes the reason you wanted a generator in the first place...
November 12, 2026 4:50 AM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ ResearcherZero,
Yup and I can think of other security issues.
On first reading it my first thought was kind of,
“How long did it take them to knock up this nonsense…”
So yup the truth real story would be nice 😉
November 12, 2026 12:25 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
@Clive Robinson
Why bother using HSMs at all if you can just knock up a script and do it in memory? What could possible go wrong with the insecure transfer of certificate authority keys using a script that was not checked to see if it performed the desired function correctly, and stored private keys in the memory of what ever systems they were using to transfer keys.
HSMs are designed to generate and store cryptographic keys without exposing them outside the hardware, but you can’t do that without the hardware security modules themselves...
November 11, 2026 11:52 PM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ Bruce, ALL,
Hot in off the press…
eMudhra gives story about why ‘4 was a Magic Number
And apparantly it was all due to being only able to take one byte to chew on…
Disgraced former CA eMudhra blames buggy script for hopelessly weak certificates
Couldn’t get the hardware it needed to generate certs, so tried to code a workaround
“Former certification authority eMudhra has claimed a buggy Python script is the main reason it issued laughably insecure digital certificates...
November 11, 2026 11:03 PM
Jimmy on AI Vulnerability Finding :
The 3% statistic reframes the whole vulnerability-management debate: if a quarter of all critical bugs ever recorded appeared in one month, then reactive patching cycles are structurally behind, not operationally sloppy. AI finding bugs at machine speed while remediation stays human-paced means the detection-defense gap becomes the primary risk surface. The proactive point in the post deserves the emphasis – triage policies and patch prioritization written for weekly discovery rates simply do not survive daily five-figure discovery rates...
November 11, 2026 9:00 PM
Artur Azevedo on Unidentified Flock Cameras in Florida :
How would a county actually find out who owns an unpermitted camera? There are only
three routes, and none of them happen by default.
Physical inspection: Flock hardware is visually distinctive enough that a code officer,
or a resident with a phone, can at least confirm it is a Flock unit and not a traffic
camera. That gets you a location and a serial number at best.
Asking the vendor: Flock knows which agency is attached to each device. But the county...
November 11, 2026 8:36 PM
Artur Azevedo on Reverse-Engineering Flock Cameras :
The detail that deserves more attention is “explicitly detects people as well as
vehicles.” Flock’s sales material – and most council presentations – rests on the
claim that these are vehicle-only sensors: no people, no faces, therefore a different
privacy calculus than a camera network pointed at a sidewalk. If the on-device model
has a person class, that premise is wrong at the architecture level, not just in edge...
November 11, 2026 7:14 PM
lurker on Three People Arrested for AI-Enhanced Impersonation Fraud :
Elsewhere, a brand new library in Coimbatore, India, was stocked with AI written books by fictitious authors, all from the same publisher at grossly inflated prices. The ~US$32M scam was picked up by an erudiite visitor detecting obvious hallucinations and dodgy ISBN nrs.
‘https://timesofindia.indiatimes.com/city/chennai/suspected-ai-written-books-phantom-authors-and-bogus-history-300cr-coimbatore-periyar-library-procurement-under-scanner/articleshow/135615876.cms...
November 11, 2026 3:41 PM
cls on New OpenSSH Vulnerability :
@Celos
The mind-set needs to be that if there are not at the very least two independent mechanisms that enforce each security property,
Um, no. Just no.
Man with 2 clocks is never sure what time it is.
Recently we’ve seen quite a number of critical defects, caused by a length here and another length over there (usually unchecked user data) and subsequent buffer underflow or overflow.
Or, vulnerabilities due to ToC/ToU gap made worse by having lengths in more than 1 place...
November 11, 2026 3:06 PM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ Bruce, ALL,
The softer the science the more goats need to be sacrificed
People might have noticed two things of recent times,
1, More and more Soft science papers are being withdrawn or recanted.
2, Increasing numbers of Soft Science papers have no or massaged data, or the data is not believable.
It’s caused a bit of an outcry with paper authors blaming each other or saying they did not have involvement in the collection or analysis of the data...
November 11, 2026 1:54 PM
lurker on Three People Arrested for AI-Enhanced Impersonation Fraud :
“It’s hard to believe how lucrative this crime can be.”
1) It is recognised as a crime;
2) The perps were caught.
That has to be positive, but it’s only an ambulance at the bottom of the cliff. These guys going to jail won’t stop others from trying the same trick. Our job is to build the fence at the top of the cliff, which is much harder. Guardrails and sandboxes are no defence against the ease of moving money to places without extraditoon treaties...
November 11, 2026 12:56 PM
Clive Robinson on Three People Arrested for AI-Enhanced Impersonation Fraud :
@ Winter,
With regards,
“I do not entertain the illusion that I could hide successfully from investigators that can spend $10M+ to find me.”
It’s actually done all the time, we just don’t get to hear about it that often.
It basically requires,
1, Knowledge
2, Training
3, Preparation
4, Planing
5, For thought
And a few other things.
The above enables you to develop a split existence and transition from one life to another before anyone starts looking for you, or more importantly has cause to...
November 11, 2026 9:39 AM
Winter on Three People Arrested for AI-Enhanced Impersonation Fraud :
It’s hard to believe how lucrative this crime can be
When such large sums are stolen from powerful people, who also lose face, it is best to be located in a pirate nest that does not extradite the perpetrators.
So, if you are not in Russia/China/NK, you are very likely to be tracked down and apprehended.
For $100M+ it is certainly worth to spend $10M+ to reclaim it.
I do not entertain the illusion that I could hide successfully from investigators that can spend $10M+ to find me...
November 11, 2026 7:49 AM
minor correction for Q on Pinpointing Location of Videos from Background Airplane Sounds :
Q: “I watched the video and was dismayed at the conclusion that the audio can’t be modified to make the technique fail.”
FWIW he was not saying that it’s impossible to modify the audio to make the technique fail.
Rather, he demonstrated why several natural/obvious methods that one would naively/intuitively expect to work won’t in fact work.
Clearly you CAN modify the audio to make the technique fail, e.g. by replacing the audio with some other audio (perhaps even re-recording your own speech from a different location later). 🙂...
November 11, 2026 7:36 AM
Clive Robinson on Three People Arrested for AI-Enhanced Impersonation Fraud :
@ Bruce, ALL,
Most new technologies bring a new looking form of fraud with them.
Eventually either people get used to it and the fraud success rate drops, or the technology goes out of fashion / use.
I used to know an accountant who was indirectly at the heart of the first major fraud by Fax Machine.
The details are dull you just need to know that a previous technology called Telex had a legal status and people assumed it carried forward with faxes (it did not)...
November 11, 2026 2:19 AM
Clive Robinson on New OpenSSH Vulnerability :
But the version of the song I give above is just about “suitable for work” because it’s had quite a bit “chopped off”.
If you are not at work then,
https://m.youtube.com/watch?v=EUthszWDgRo
And yes in my past, I used to sing it for “free beer” in a pub on the edge of South West London…
November 11, 2026 2:08 AM
Clive Robinson on New OpenSSH Vulnerability :
@ Scott,
Hark,
“What light through yonder window…”
With regards, your statement
“Of COURSE *sshole is meant to invoke the curse word for your rear end. Of COURSE it is. Are you kidding?”
At my comment of,
“The name given to the vulnerability of “*sshole” is not supposed to be seen as “rude”.”
It’s why I then gave the explanation you can say to your boss etc with a perfectly straight face… But I guess it’s lost on some people...
November 11, 2026 12:18 AM
Mr Obvious on Bloatware in Aftermarket Auto Parts :
De-bloating an Android device using ADB over wifi/USB is not news. It’s nice that Claude can do it but people have been doing this kind of thing for ages. Isn’t this the best/worst part about Android – that it’s open/hackable?
November 10, 2026 11:08 PM
Anonymous on New OpenSSH Vulnerability :
@Smooth Operator,
exactly that.
I remember Deep Throat
https://threats.kaspersky.com/en/threat/Backdoor.Win32.DeepThroat/
November 10, 2026 11:00 PM
Smooth Operator on New OpenSSH Vulnerability :
@Scott • November 10, 2026 1:24 PM,
If he was even a tiny bit familiar with the culture of those circles then he’d known that the hacking community culture prioritizes the use of terms like that, wherever and whenever possible, and examples thereof are plenty out there.
Just google “deep throat” or “back orifice” (just to name a couple of numerous examples).
Of course he knew this. Just watch him get smart with us. HE KNOWS EVERYTHING...
November 10, 2026 10:44 PM
Jimmy on AIs Hacking the Law :
The scale point in the Ringe excerpt is what changes the game: courts have always absorbed the occasional clever exploit through interpretation, but interpretation as a defense assumes a trickle, not a flood. Cheap legal reasoning turns isolated cleverness into systematic search, the same way cheap automation turned isolated API quirks into abuse at scale. Rule-governed systems tend to harden only after the flood becomes visible. I run a small independent tracker for AI coding-agent usage-limit reset windows (...
November 10, 2026 6:54 PM
anonymouse random on New OpenSSH Vulnerability :
SSH has become too complex.
Cut loose the legacy crap.
Smaller attack surface. <
SSH is very-widely-used legacy code. I believe that it has not been much changed since ~2006.
This bug arose from a contract change in a library used by a few SSH implementations. It shouldn't need saying that library publishers shouldn't change their libraries' contracts in any way that breaks backward compatibility...
November 10, 2026 5:55 PM
Winter on Pinpointing Location of Videos from Background Airplane Sounds :
@Clive
All without creating new differentiating signals to trip up on…
It’s an arms race. Technology can identify people on their voices alone. There are actual challenges organised to spoof automatic speaker recognition, and challenges to foil the spoofers.
In the end it is a matter of resources. And technology tends to make things cheaper over time.
What yesterday was exclusively the field of the TLAs, is today the weapon of organized crime, and tomorrow the tool of the script kiddies...
November 10, 2026 5:50 PM
Uh, Mike on New OpenSSH Vulnerability :
SSH has become too complex.
Cut loose the legacy crap.
Smaller attack surface.
November 10, 2026 5:05 PM
Clive Robinson on Pinpointing Location of Videos from Background Airplane Sounds :
@ winter,
You make the statement of,
“In general, what can be measured, can also be faked.”
Without qualification.
Yes you could make all manner of measurements but two questions arise,
1, Have you made all the necessary measurements required?
2, Are there the resources to fake all the results sufficiently?
All without creating new differentiating signals to trip up on…
November 10, 2026 3:59 PM
Winter on Pinpointing Location of Videos from Background Airplane Sounds :
@Clive
Primarily because the fake signals will always have some kind of distinguisher you won’t be aware of or can not account for.
In general, what can be measured, can also be faked.
The way the bad guys get around DNA evidence is by emptying a full hoover bag of dust over the crime scene. Fingerprints can be copied and left at the scene of a crime. Blood and DNA can be removed with bleach, or added from storage...
November 10, 2026 3:49 PM
Winter on Pinpointing Location of Videos from Background Airplane Sounds :
@Clive
Primarily because the fake signals will always have some kind of distinguisher you won’t be aware of or can not account for.
In general, what can be measured, can also be faked.
The way the bad guys get around DNA evidence is by emptying a full hoover bag of dust over the crime scene. Fingerprints can be copied and left at the scene of a crime. Blood and DNA can be removed with bleach, or added from storage...
November 10, 2026 3:47 PM
Blinky on New OpenSSH Vulnerability :
I think it is probably a “double-entendre”.
November 10, 2026 3:42 PM
lurker on New OpenSSH Vulnerability :
ISTR sometime in the ’80s being given a shell account on some flavour of *nix, and being told the maximum size my password could be was 55 (or 56?) characters. So apparently that flavour of *nix had got the problem sussed 40 years ago. People never learn from history and forever repeat it …
@freedom
When there’s nobody in charge, there’s no oversight, the dots, or *s, don’t get joined.
...November 10, 2026 3:36 PM
anonymouse random on New OpenSSH Vulnerability :
musl C library’s crypt() function returns * when password hashing fails
That wasn’t initially the case, but the developers changed the crypt() function’s contract (a king-sized no-no for library functions), which allowed the exploit:
“The crypt() function, which generates password hashes for comparison with stored values, can return NULL if it can’t compute a hash. Applications that fail to check that result may crash. Musl libc’s developers instead chose to return a single character for certain failures, and they picked the asterisk.” (From The Register ; search “musl” . Link omitted because automod dislikes it. Can someone fix automod?)...
November 10, 2026 2:39 PM
freedom on New OpenSSH Vulnerability :
ssh accounts configured with * to block password logins.
musl C library’s crypt() function returns * when password hashing fails
an attacker can submit an oversized password and cause OpenSSH to compare the returned * against the account’s stored *, resulting in a successful login.
Ha ha. The field of “computer security” is just not serious…
November 10, 2026 12:52 PM
Celos on New OpenSSH Vulnerability :
And that is exactly why we need redundant, resilient security. In the case at hand, there should have been a length-limit before that is independent of the C library and enforces the correct behavior. The mind-set needs to be that if there are not at the very least two independent mechanisms that enforce each security property, then it may well be regarded as broken. And, obviously, that needs to be the absolute minimal standard for anything that is Internet-reachable...
November 10, 2026 11:22 AM
Clive Robinson on New OpenSSH Vulnerability :
@ ALL,
The name given to the vulnerability of “*sshole” is not supposed to be seen as “rude”.
The “sshole” is actually a contractin of “SSH hole”. The lead “” is part of what makes the vulnerability work. And is based on the *nix password file, where the “” was placed in the password field to indicate an account was not to be loged into from the login programe
The reason for the lead “*” is it was not a valid character in the password field and thus could be used for “in band signalling”...
November 10, 2026 9:04 AM
Andy on Friday Squid Blogging: New Jurassic Squid Predecessors :
Formal proof of algorithm correctness should be of significant interest to @Bruce, particularly given the example they pose. Proving the security of a cryptographic algorithm is a big deal. I recall attempts to do this with avionics code back in the 1990’s with only modest success.
November 10, 2026 7:16 AM
Privacy on Pinpointing Location of Videos from Background Airplane Sounds :
Clive, I used to sell fault recorders and protective relays to electric utilities and repair those.
You are correct that arcing can be localised. Broken conductors can be localised. Mains frequency can be localised. Fault data was timestamped.
The utilities used the timing to determine distance-to-fault so a repair crew could be dispatched to the fault location.
The above equipment triggered on arcing, overcurrent, voltage, loss of load and frequency getting out of narrow limits. Inputs came from CVTs and CTs...
November 10, 2026 1:29 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
Poor security practices by an Australian automotive firm left vehicles remotely accessible.
https://shubs.io/unlocking-any-car-from-x-brand-in-most-of-the-world/
The mobile apps used by the unnamed automakers had some serious problems.
https://samcurry.net/forgerock
November 10, 2026 1:17 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
@Clive Robinson
At least The Lost Generation had World War I to look forward to.
–
Get-AppxPackage *Paint* -AllUsers| Remove-AppPackage –AllUsers -verbose
November 10, 2026 1:11 AM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
Attribution is hsrd
but not if you’re French: Iran melted their fridges (set fire to a few), and CERT-FR has isued a report on it.
November 10, 2026 12:34 AM
Winter on AI Security :
MIT has succeeded in making one AI directly edit the weights of another. Thereby making a targeted change in the behavior of the second AI.
‘https://www.mindstudio.ai/blog/gpt-5-astra-weight-space-programming
If that could be developed, that’s a big IF, it could be used to build Asimov style security laws into AIs.
The down side is that it also allows to compromise any AI. And I suspect there will be a proof that you cannot always ensure an AI will not have been corrupted...
November 9, 2026 4:21 PM
Clive Robinson on AI Security :
@ All,
Apparently Mr Nvidia has not been keeping it holstered… And for the past year or so been “playing away from home”,
“It is entirely possible that Jensen Huang may have passed state secrets from Trump to Alice Huang. Certainly Jensen Huang was allowed to continue to lobby the President around sensitive U.S. policy matters around chips and China. The incoming Congress is certain to investigate.”...
November 9, 2026 3:54 PM
DreamMMer team on AI Security :
The split into security of AI, security with AI and security from AI is a useful frame, especially for teams building creative tools. For a generative image and video product, ‘of’ means protecting models, prompts and user uploads, ‘with’ means using agents to catch misuse, and ‘from’ means limiting abuse like non-consensual imagery. I’m from the DreamMMer team (an agent-driven image and video workspace), and we find it helps to keep those three questions separate when deciding what an agent may do on its own. Looking forward to the talk...
November 9, 2026 3:44 PM
Clive Robinson on AI Security :
@ ALL,
For those that are a little skeptical of some of what I say above,
Have a read of,
AI Agents Are Testing the Limits of Zero-Trust Security
As autonomous AI agents work together, verifying their actions, enforcing security policies, and maintaining accountability becomes increasingly difficult
“Several teams around the globe are leveraging AI by creating a task force of agents managed by an orchestrator and creating applications to support risk assessment, client profiling, match making, and information yield with a slight overview of decision making. One of the important components for these agents is to ensure that the logs are managed and historic relevance is not being used to avoid skewing towards same decision making. Thus, the system is expected to keep checkpoints in place where agents are prompted to delete such historic relevance...
November 9, 2026 3:31 PM
Clive Robinson on AI Security :
@ Bruce, ALL,
With regards,
“AI security can be divided into three broad topics”
Of Security “of / with / from” AI.
But each of those also needs to be considered in “Past / Present / future” tenses.
As we can tell from the Hugging face and subsequent incidents, to few researchers appear concerned about Past and Present tenses, and not sufficiently clued up for predicting yet alone acting on the future tense, which is quite worrying as the appear locked into a...
November 9, 2026 1:14 PM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ ResearcherZero, ALL,
With regards adolescents and your observation that they are in effect “cast adrift” thus easy prey for,
“There are some who are willing to offer a sense of certainty and belonging.”
It’s actually worse than most realise…
Back at the begining of this century there was terrorists arising from the Middle East. The reasons why there was and still is such terrorism can be laid fair and square at the doors of the US State Dept and those setting policy even against the directions of the US Executive and of other US politicians having a majority in one of the houses…...
November 9, 2026 12:40 PM
Joshi on AI Security :
“The World is built in a wall that separates kind. Tell either side there’s no wall, you’ve bought a war. Or a slaughter.“
November 9, 2026 12:37 PM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ StephenM, lurker, ALL,
With regards Palantir and the NHS.
I’ve made comment before about Palantir and it’s,
“Pushing drugs at the school gate”
Style business practices.
The NHS records are supposed to be “siloed” from all other UK Central and local Government agencies / entities and any commercial organisation. As they are “special category”. That is they are treated as “confidential” or higher equivalent in secret “documentation classification” terms...
November 9, 2026 12:28 PM
Rontea on AI Security :
Reflecting on this today, the 37th anniversary of the fall of the Berlin Wall, I can’t help but think about how barriers—physical or digital—shape our world. Just as the Wall symbolized division and control, the challenges in AI security remind us that our defenses and safeguards can either isolate innovation or protect freedom. Our work in securing AI is about making sure that these new systems empower rather than imprison, building trust instead of walls...
November 9, 2026 11:15 AM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
@StephenM
“It won’t be easy, but next spring the job of clawing it back must begin”
Oh dear. The innocents believe that Palantir knows and respects the right to be forgotten.
November 9, 2026 8:04 AM
gheese on AI Security :
Looking forward to the recordings! I believe they’re not available yet, but I’ve bookmarked a couple of slide-sets already.
November 9, 2026 5:57 AM
StephenM on Friday Squid Blogging: New Jurassic Squid Predecessors :
@Clive
“British people don’t trust Palantir – it’s not too late to reverse the NHS’s dependence on it”
November 9, 2026 3:51 AM
Anonymous on Friday Squid Blogging: New Jurassic Squid Predecessors :
Even the CEO of Nvidia has trouble controlling his impulses.
Google “Fang Fang” for semi-related information.
November 9, 2026 3:23 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
A generation failed by society, policy, industry and institutions, without direction or any sense of security, acceptance and understanding. Essentially abandoned and unsupported.
There are some who are willing to offer a sense of certainty and belonging.
Adolescents experience a period of seeking out absolute certainty. This can expose them to dubious sources of information that offer certainty. The prefrontal cortex matures more slowly than the emotional center of the brain. This part of the brain governs impulse control and rational judgement...
November 9, 2026 12:50 AM
ResearcherZero on Friday Squid Blogging: New Jurassic Squid Predecessors :
Even the CEO of Nvidia has trouble controlling his impulses.
While it may be difficult to take responsibility for your own actions, and you may even perhaps believe that your own behaviors are outside of your control, preparation, planning and a discipline will reduce the risk of exposure and compromise. What is, and what is not outside of one’s control, can be a choice if reasonable steps are taken to minimize the risks, rather than defer daily routines and responsibilities to the domains of convenience and denial...
November 9, 2026 12:50 AM
lurker on Friday Squid Blogging: New Jurassic Squid Predecessors :
@Clive Robinson, ALL
I started to read the FCC documents, and my eyes glazed over. It seems like the US has forgotten how to make anything except laws, and not very good laws at that. Some sources say that 82% of all “FCC approved” devices last year were approved by labs in China. What were they smoking when they decided the US, or Elbonia, or who would be able to pick up that work and run with it, inside 12 months?...
November 8, 2026 10:34 PM
Clive Robinson on Friday Squid Blogging: New Jurassic Squid Predecessors :
@ lurker, ALL,
With regards the FCC and testing radio equipment it’s actually quit involved.
Officially the party line is about “reciprocal testing” but the reality is the Executive has designated much communications equipment such that it can not be “Exported to China” (kind of like the GPU issue).
So there are no US companies wanting to export to China because they are not allowed to do so. Which means no US company wants to have equipment tested to China’s standards. So there is no requirement for reciprocal testing…...
November 8, 2026 10:05 PM
freedom on Friday Squid Blogging: New Jurassic Squid Predecessors :
new FCC Rules ban foreign testing labs (mostly Chinese)
Well, you know, the U.S. is the supreme champion of “free trade” as long as they monopolize all trade.
Sidebar photo of Bruce Schneier by Joe MacInnis.
@ Ismar, ALL,
With regards,
The idea of “vibe-coding” was to,
“Get the LLM to ‘cut the code’ and do the syntax correct typing.”
So a human can,
“Wave the hand to ‘stitch patterns together'”
The problem is that whilst generic patterns may be semi-secure the interfaces and communications between patterns are where the security failings all to frequently occur...