The South Pacific Regional Fisheries Management Organization (SPRFMO) needs to regulate squid fishing in the South Pacific.
As usual, you can also use this squid post to talk about the security stories in the news that I haven’t covered.
Clive Robinson • May 23, 2026 8:14 AM
@ J van Prooijen,
With regards,
“So if you use AI for age verification for access management it should be trained with images of dishonest people.”
Hmm… And thereby hangs another problem…
As you note there are,
1, Honest people
2, Dishonest people
But there are at least another two groups,
3, Dishonest people that are assumed to be honest.
4, Honest people that are assumed to be dishonest.
The latter group will either complain if they can (almost impossible these days). Or they will simply “walk away”…
But this is where it gets interesting… Your proposal will “train against them” and that is certainly “unlawful discrimination” in most places in the West. But it degrades the reliability of your system.
The flip side of this is the third group, your system will put them in the “Honest people” group thus train up on them and their tricks thus degrade reliability yet again.
Due to the fact that the system is in effect “static defence”, those in the second group will by a simple “evolutionary response” process of “active attack” work out how to get into the third group. Which when successful puts them into the desired first group. As bio-metrics are really bad at anything[1] this process is really “childs play” to “find a way”.
[1] Tests have shown that video based biometrics are really quite ineffectual with error rates only being small in small groups of candidates. Large groups have large error rates and are quickly rendered more or less useless. With law suits starting to occur more and more frequently.
Jel znades tko ti je otac? JA. Pitaj Mamu • May 23, 2026 9:17 AM
@ snortin my future • May 22, 2026 7:20 PM
Jbm ti mamicu usranu, kaj te nije abortirala fkn muddy $l1m3.
Jealous Guy Aren't We? • May 23, 2026 9:29 AM
@@ snortin my future • May 22, 2026 7:20 PM
Look, “bro” or “bruh” or boohznyun monkey….
I cant help you out here that your momma, your sister, and your GF are all dying to get into my pants… not my fault…talk to them..
Signed:
The “UNCATHCABLE” Fake French spy.
kandu • May 23, 2026 3:41 PM
Has anyone hear heard of turning the video output of into a wifi transceiver?
I have wifi signals coming out of my video cable. It’s a good displayport cable and it looks very much like wifi signals in the 5.0 to 5.8 ghz range. It is not a data rate harmonic or anything like that but rather wifi signals. The computer I’m using has problems and I have to re-install the os every couple of months. It isn’t that old maybe a year. I’ve remove the internal wifi adapter but its pretty persistent.
5.049 ghz
5.197 ghz
5.342 ghz
5.519 ghz
5.667 ghz
5.710 ghz
5.720 ghz
lover's lane (binary edition) • May 23, 2026 4:06 PM
@ kandu,
Using HDMI EMI for fast wireless data transfer
https://www.windytan.com/2023/02/using-hdmi-radio-interference-for-high.html
Meanwhile, to my secret admirers:
Now I’ve got feet
and you’ve got feet
Let’s start to walk where the lovers meet
Let’s go
Times a wastin
Clive Robinson • May 24, 2026 9:46 AM
@ Bruce, ALL,
AI knackers Google Search more than Bing…
Or so this journalist thinks…
You can no longer Google the word ‘disregard’
“Earlier this week, Google rolled out a completely new Search experience, foregrounding AI summaries and kicking the traditional “10 blue links” far down the page. But the sheer scale of Google Search means there are lots of edge cases that the company doesn’t seem to have considered.”“
https://techcrunch.com/2026/05/22/you-can-no-longer-google-the-word-disregard/
Basically Google’s AI interface looks crap to the journalist. Whilst I can not say anything about Google having not used it for years, I can say that Microsoft’s Bing AI via DuckDuck is nothing I would want to have around my searches…
Which is why the journalist s ultimate paragraph made me smile about my choice to dump Google some years ago…
“I have been a professional tech journalist for nearly 15 years, and before today, I cannot think of a single time when a Bing search result was more valuable than the Google equivalent. There really is a first time for everything!“
As for Google and “disregard” being effectively banned I think most here will thing in terms of “AI Guiderails” and “Prompt attacks”.
reap the whirlwind • May 24, 2026 11:00 PM
@ Daisy The Ruler of StingRays,
No daisy at all, it would seem. ^_^
Anonymous • May 25, 2026 5:59 AM
kids faking age verification:
i asked my kid about this. he smiled and stated “if this does not work, just use a virtual cam an a game avatar”.
not sure if that’s realistic, but there seem to be many ways to cheat technology.
remember: if you think, technology solves your problem, you don’t understand technology and you don’t understand your problem.
i am always amused when marc z. states in congress hearings “we are already trying so hard to solve the problem”.
ever tried to convince an investor with this statement?
they would probably say: “solve the problem or we’re out.”
if governments were serious, they would do the same.
Clive Robinson • May 25, 2026 9:58 AM
@ Anonymous, ALL,
With regards,
“but there seem to be many ways to cheat technology.”
Any “tangible physical object” based identifier Technology is easy to cheat at the point it becomes an “intangible information object”.
Put simply between the “tangible physical object” and “intangible information object” is a sensor doing the translation from physical attributes to information values.
But by necessity there has to be,
“A gap twixt physical object and sensor, in which the devil can play.”
In this gap just about any “devilish fun” can be had. It’s one of the reasons I rank nearly all bio-metrics as “bogus nonsense”.
I’ve been through this in more depth in the past, but the simple fact is,
“There can not be a single security chain from “tangible physical objects” to “intangible information objects”
Find the end of one security chain and thus the “gap” to the begining of the subsequent security chain and you have control of the “translation” from one object type to the other type.
The fact “kids” appear to be faster on the uptake to this than Politicians and even Security Gurus, says a lot about why all ID systems will always have an inbuilt failure mode… Then of course there are the shills and Snake oil salesmen looking to capitalise on such ignorance for their profit…
Anonymous • May 25, 2026 5:51 PM
@Clive, All
so, what do you think would happen if governments grant permission for a platform to exist only on the basis of functioning age verification?
You Always Reap What You Sow • May 26, 2026 12:40 AM
s h or t u r l . a t / x9q9k
BOISE, IDAHO:
The following Boise Police Department Officers (current or former)
Matt Hudson,
Edward Pieczonka,
Trent Schneider,
Chad Wigington,
Damir Subasic,
as well as their two former, (both TERMINATED) Chiefs:
Bill Bones and
Ryan Lee,
AS WELL AS THE @D@ COUNTY PR0SECUT0RS:
Jan Bennetts and
Brittany Ford,
as well as the Public Defenders:
Jonathan D. Loschi and
Kendra Nagy,
as well as the PRIVATE LAWYERS THAT THE INNOCENT AMERICAN CITIZEN HAD RETAINED:
these private lawyers COLLECTED THE MONEY FROM AN INNOCENT MAN TO DEFEND HIM
BUT THEY DID NOT EVEN FILE AN APPEAL EVEN THOUGH THE MAN WHO PAID THEM SPECIFICALLY
ASKED THEM TO CLEAR HIS NAME OF ALL FAKE CHARGES WHICH HAVE DESTROYED HIM AND
HIS FAMILY.
Raymond D. Schild
John Prior and
Charles Crafts
ALL OF THE ABOVE CREATURES
have destroyed an INNOCENT MAN AND HIS FAMILY BY WITHHOLDING THE
EVIDENCE OF AN ATTEMPTED MURDER ON THE ACTUAL VICTIM BY THE FAKE VICTIM
husein kurchich.
Exposing the COVERUP RIGHT HERE BELOW:
s h o r t u r l . a t / x9q9k
GOD BLESS ISRAEL!
Clive Robinson • May 26, 2026 1:41 AM
@ Anonymous, ALL,
With regards,
“… on the basis of functioning age verification”
There is no such thing, nor can there be…
There is a “tangible physical object” and an “intangible information object” and contrary to what every idiot politician thinks / gets told they can not nor never will be verifiably linked.
To see why consider the simple problem starts with the primary document your “birth certificate” do you see any way to link it to you reliably and without some one simply taking it or getting a copy to become you?
No, because you are a “tangible physical object” the birth certificate is another “tangible physical object” on which an “intangible information object” is somehow impressed/modulated.
Take an official “watermarked” and “security threaded” page and then print on the information.
The problem is that nearly all modern methods of putting ink on such a page are generally reversible (the ink “sits on the page” rather than impregnates irreversibly into the fibers with chemical changes).
It’s why in some places they actually use a laser to burn in the writing into the page, because it can not realistically be reversed.
But the problem is that there still is no “secure link” between the supposedly “secure paper” physical object and the information encoded in the printing method used for the information object…
In theory the page could be “uniquely identifiable” and that somehow used to “sign” the information.
The problem is what verifies the information?
Yup nothing…
Lets say they “use your blood” as the ink…
The thing is it still does not verify the information nor attest to it being correct.
Because what ever process is used, you will always “find a gap” by which false information can be made to look physically trustworthy (ie all credentials can be forged).
As Dame Stella Rimington first female head of the UK National Security Service MI5 observed,
“My angle on ID cards is that they may be of some use but only if they can be made unforgeable – and all our other documentation is quite easy to forge.
If we have ID cards at vast expense and people can go into a back room and forge them they are going to be absolutely useless.”[1]
The point is “the gap” I refer to will always be present when you try to turn a “tangible physical object” into an “intangible information object” or vice versa. It’s impossible to get around. Therefore all ID as “recorded information” will be forgeable.
[1] See this BBC article,
http://news.bbc.co.uk/1/hi/uk_politics/4444512.stm
And note that it goes back to a time when the then UK Prime Minister Tony Blair was trying to “pay off” companies and organisations he had received “benefits from”. The much more recent failure political failure in the UK on ID Cards was to not just pay-off “Tony’s obligations” but also to enrich one of his children… Proving if needed that Tony and cronies still run the political party agenda on behalf of “vested interests”.
Clive Robinson • May 26, 2026 2:26 AM
@ Anonymous, All,
Don’t let people disturb your train of thought…
Whilst I was tapping in my above I got disturbed by the phone.
The result was that I left a chunk of info out of the foot note.
Yes the recent ID Card motion was a failure, and various reasons have been given,
However the one many blaim was the story about Tony Blair and his Son’s company, it’s since been “debunked”,
https://www.no2id.uk/2025/multiverse-and-euan-blair/
But at the time it was widely reported as factual and made a lot of noise.
It’s just been in the news again,
As various people want to “kick the truth into the long grass”.
The last paragraph puts a little hint as to why in the air,
“In other words, MPs aren’t entirely convinced the same government that brought Britain some of its more memorable public sector IT disasters should be trusted to build national identity infrastructure.”
And Tony Definitely had his thumb in many of those “IT disasters”…
Which others such as Boris Johnson gave away to Palantir, a US company that want’s to replace all the “Intelligence Analysts”, “Detectives”, and similar with AI…
Noam • May 26, 2026 2:28 AM
@You Always Reap…
That is terrible what happened to you- you should ask Trump to help you . I am sure American people are foremost on his mind .
Clive Robinson • May 26, 2026 7:54 AM
@ Bruce, ALL,
Politically motivated surveillance by UK Met Police.
A UK FOI request was made against the UK Metropolitan Police for their requests to “Mobile Virtual Network Operators”(MVNOs)[1].
What it threw up was a decidedly “politically motivated” policing policy,
London’s police asked Big Tech for comms data over 700,000 times last year
A Freedom of Information Act request shows the extent of the surveillance
“London’s Metropolitan Police – the UK’s largest police force – asked tech companies to give officers access to private communications data over 700,000 times in 2025 alone, according to figures obtained by The Register under the Freedom of Information Act.
These statistics expose the monitoring of everyday platforms like takeaway delivery services, and also show a massive surge in the force’s surveillance of the users of low cost MVNO LycaMobile. Additionally, our FoI exposed the acquisition of data from encrypted messaging services designed to offer privacy.
Since 2024, the Met says that it has obtained communications data (CD) from Proton’s privacy-focused mail service users 139 times. CD is not messaging content, but metadata. In Proton’s case, this could include account payment details and, in some instances, IP addresses.“
The rest of the article quickly becomes “eye raising” and the UK Metropolitan Police are obviously being “less than candid” if not basically dishonest.
[1] A mobile virtual network operator or MVNO is an organisation that sells access to mobile phone networks belonging to Network Service Operators that hold the licences to Operate the Radio side of such networks. These days the licence holders rarely if ever own the actual equipment masts or other infrastructure they simply rent or lease back in various ways.
The advantage of an MVNO is they can offer niche products that are often less expensive or have specific advantages with regards communications including ways to pay in cash or other anonymous way. Some even offer “virtual/one time numbers” which like “virtal/one time credit cards etc help protect “vulnerable people” in all sorts of ways against those who wish them harm.
Anonymous • May 26, 2026 2:26 PM
@Clive, All
Sorry to be so unprecise.
you wrote:
“… on the basis of functioning age verification”
There is no such thing, nor can there be…
I understood this fact from your first post already. only governments seem to insist in not understanding.
so, instead of accepting failed attempts for age verification, why don’t governments insist in the result.
make the corporations liable and attach a large price tag. no excuses accepted.
if coprorations risk to get sued for every single proven case of non functioning age verification check, they figure out the necessary consequences by looking at the numbers.
same as investors would do with their requirements.
lurker • May 26, 2026 3:42 PM
Shock headlines in some of the tech press who should know better:
Huawei to make 1.4nm chips
Actually Huawei themselves claim only
“By 2031, the high-end chips HUAWEI designs based on the τ Scaling Law are expected to feature a transistor density that is equivalent to 14 Å (1.4 nm) processes.”
https://www.huawei.com/en/news/2026/5/ieee-iscas-tau-scaling
The trick is to layer chip dies on top of each other to acheive a higher transistor density per chip area. By stacking the right parts over each other cross-“chip” transit times can be drastically reduced. Huawei call this Logic-folding and suggest a tau-law to replace Moores law.
El Reg notes some scepticism
https://www.wired.com/story/us-law-enforcement-warns-of-anti-tech-extremism/
anti-tech? or anti-anti-constitutiin?
there’s probably a shadow docket rationalizing all of this somewhere around here.
the wild wild hunt! and dawn dies at dusk! • May 26, 2026 7:54 PM
@ Noam,
That is terrible what happened to you- you should ask
Trump to help you . I am sure American people are
foremost on his mind .
What happened to me? You’re no daisy, my friend. No daisy at all! 😀 And who said I was American?
Clive Robinson • May 26, 2026 9:43 PM
@ lurker, ALL,
With regards the article,
“The trick is to layer chip dies on top of each other to acheive a higher transistor density per chip area.”
It has been tried before as a way to improve “yield” but the reality is,
1, Has high heat death issues.
2, Has mechanical issues.
3, Has “tape out” issues.
4, Has alignment issues.
5, Has other production issues.
6, …
Whilst I’m reasonably certain solutions to these issues are possible, the question of “When?,” arises. To me 2031 feels to optimistic.
Clive Robinson • May 26, 2026 9:45 PM
@ lurker, ALL,
With regards the article,
“The trick is to layer chip dies on top of each other to achieve a higher transistor density per chip area.”
It has been tried before as a way to improve “yield” but the reality is,
1, Has high heat death issues.
2, Has mechanical issues.
3, Has “tape out” issues.
4, Has alignment issues.
5, Has other production issues.
6, etc, etc.
Whilst I’m reasonably certain solutions to these issues are possible, the question of “When?” arises. To me 2031 feels shall we say “a little optimistic”.
Clive Robinson • May 27, 2026 4:23 PM
@ ALL,
AI slop destroying reputations credibility and the Internet.
A little while back I noted that there was “investment advice” video that was “supposedly” from a well known finance industry personality,
And whilst I was not sure, some things made me think it had been AI Generated[1].
It turns out others of this blogs readers felt the same…
Looking at the video provider showed they were putting up videos at an alarming rate of more than six or ten a day.
But also confirming things within 12hours the poster got hit by a copyright strike and had all videos taken down, which kind of confirmed it was not legal content.
Well…
Turns out it’s becoming a major problem the latest I’d noticed a week or so back was Richard Feynman video’s were popping up. That were just wrong in oh so many ways.
Well it turns out I’m very far from being the only one spotting AI slop just from the thumbnail and title…
Thus people have produced their own videos on this kind of AI slop, how to spot it and more importantly what to do about it,
https://m.youtube.com/watch?v=A2_8199CL1I
This year will probably go down in the history books as,
“The year the internet got Enshittified by AI slop!”
Realistically, this will grab “advertising revenue” taking income from genuine content providers, and advertisers will stop providing money for fake audiences etc.
I’ve already seen this pull back start in “Amateur Radio” and other semi technical subjects. Where the response is to make “genuine slop” such as 12hours of a “channel scanner” jumping on US first responder frequencies in “some small town etc” and even military frequencies sent out to aircraft that form “first line response” etc. As an example the latest of many such nonsense is,
https://m.youtube.com/watch?v=zE8Xze-t7zg
The question that obviously arises is,
“Has the financial models of big US Tech / social media caused the Internet to be killed off by AI Slop?”
Me I think the answer is almost certainly “yes” which suggests that things have to change, but will they in time?
[1] The two things that caught my attention was the thumbnail and looking at the video lead in it appeared to be made of very short 8 second video clips “camera jumping” stitched together, but with image and voice breaks synchronised almost like the beat of a metronome, which rarely happens in real life and certainly not for a half hour or more. Secondly the lack of facial feature changes and even less body movement. But interestingly what was being said was in effect coherent and sufficiently believable to keep people watching.
lurker • May 27, 2026 10:31 PM
@Clive Robinson
re your list of issues with layered chips
Yes, as a hardware guy all those had crossed my mind. I half expected Huawei might have found a way to multilayer etch/deposit, since they are still claiming only 7nm parts. There’s obviously more to this story.
lurker • May 27, 2026 10:35 PM
@Clive Robinson
re YT “zE8Xze-t7zg”
This is a private video. Please sign in to verify that you may see it.
Clive Robinson • May 28, 2026 3:19 AM
@ lurker, ALL,
With regards,
“This is a private video. Please sign in to verify that you may see it.”
Hmm, I just checked (06:45 UTC) and yes it now requires a “sign in” that it did not have before…
It was a video of two hand held scanners upright side by side on a table with a camera showing the screens…
This one is similar but is “ham radio” rather than “emergency responder”. And a little more “professional” in that it has a one and a half minute lead-in before just about 12 hours of “PC based SDR software screen”,
https://m.youtube.com/watch?v=fwgfTSPqOYk
He has several such on his “channel”,
https://m.youtube.com/@KG6NLW/streams
There are several other such “channels” they are not difficult to find.
Some indicate that it’s “against the algorithm” others well no reason given. Such as the earlier versions that are still going that are the many,
“study with me”
Channels, and bird table cams, nesting box cams, deer feeder cams, etc, etc. Most of which appear “fully automated” unlike the “study with me” ones.
Clive Robinson • May 28, 2026 11:28 AM
@ lurker, ALL,
Getting back to,
“Enshittification by AI slop”
It appears that a lot more are noticing it as well…
YouTube has indicated it’s going to use AI and other algorithms to catch AI Slop and label it…
Over on Hacker News they have a comment thread on the YouTube announcement,
https://news.ycombinator.com/item?id=48299753
That was quickly getting lengthy…
Clive Robinson • May 28, 2026 12:41 PM
@ Bruce, ALL,
Caims that US troops can be targeted by commercial location data
Whilst this “should be obvious” to anyone who can think, it appears to be a bit of a surprise to law makers and others.
Put simply if you get your entire populous surveilled by commercial companies that then “give it for free” to certain Government agencies.
Then you should expect not only will the companies be “inclusive of everyone” but also they will make the data available in the marketplace that arises. Thus anyone with sufficient money can in effect locate anyone in their “population group” of choice.
U.S. says troops were targeted with location data, as senator warns ad industry is a ‘national security threat’
“The U.S. Department of Defense has confirmed that adversaries have targeted and surveilled serving military personnel on the battlefield using commercial location data, the latest demonstration of how information collected from phones and computers can be abused to track and target individuals.
In a letter shared by Sen. Ron Wyden with TechCrunch, U.S. Central Command said it was aware of hostile actors using purchased location data to track U.S. servicemembers.
Clive Robinson • May 28, 2026 12:45 PM
@ ALL,
Opps, left link to Tech Crunch out of my above,
(Second attempt to post due to 429 error…)
Clive Robinson • May 28, 2026 12:55 PM
@ ALL,
Opps, left link to Tech Crunch out of my above,
(Second attempt to post due to 429 “rate limit” error…)
Clive Robinson • May 28, 2026 8:05 PM
@ Winter, Usual Suspects,
Based on previous discussions you should find this of some interest,
https://phys.org/news/2026-05-dark-energy-equation-mathematicians-standard.html
One statement in it did make me smile,
‘Study corresponding author Blake Temple, a distinguished professor emeritus of mathematics at UC Davis, compared the standard cosmological model to a pencil standing on its tip.
“All the forces are in balance when a pencil is standing on end, so it is a ‘solution of the equations,'” he said. “But it’s unstable. Any breath of air and it falls away.”
The mathematics, Temple said, prove that Friedmann spacetimes—mathematical models that govern cosmic expansion—are unstable at both small and large length scales at the Big Bang, making it the most unstable solution of all.
“Unstable solutions in physics and science are considered not physical,” Temple said. “You’ll never observe them in nature.”‘
It is a principle of much broader basis and applies in fundamental ways to many things including security.
It’s something we’ve discussed in the past.
Winter • May 28, 2026 10:12 PM
@Clive
‘Study corresponding author Blake Temple, a distinguished professor emeritus of mathematics at UC Davis, compared the standard cosmological model to a pencil standing on its tip.
Dark energy and matter are the odd ones out in physics as they are not part of the standard model of physics, ie, there are no particles/fields associated with them.
They are both incorporated in cosmology because general relativity does not explain the observed behavior of space and mass at the largest scales.
Hence the search for extensions/adaptations of general relativity that could better model the behavior at these largest scales without adding mass and energy outside of the standard model.
Such adaptations are not easy to conceptualize. Dark matter and energy play a role when gravity is extremely “weak”.
This is the part of theoretical physics where mass and energy, ie, the stress-energy tensor in GR, create the very space and time it lives in. So, when/where the influence of mass/energy is very weak, space and time might start to behave in “odd” ways.
Odd ways that express themselves as if there is more mass than accounted for and there is pure energy in empty space.
All will be solved when Quantum Gravity is fully formulated, which will integrate Quantum Mechanics and General Relativity.
Or so we are told. However no one knows yet how quantum gravity will look and what surprises still lurk in empty space.
capsule • May 29, 2026 12:10 AM
CIFSwitch Vulnerability Exposes Some Linux Distros to Local Root Access
https://linuxiac.com/cifswitch-vulnerability-exposes-some-linux-distros-to-local-root-access/
The flaw affects the boundary between the Linux CIFS client and cifs-utils, allowing local root access on some systems.
Clive Robinson • May 29, 2026 8:04 AM
@ capsule, ALL
CIFSwitch Vulnerability walk through and future issues
I read about CIFswitch at,
https://heyitsas.im/posts/cifswitch/
For various reasons it makes interesting reading but two main points arise,
1, How to use DNN based systems to find vulnerabilities by what is in effect the first stage of a “McCabe Analysis”[1].
2, The issues of insufficient security regulating historically brought forward from resource limited times.
The McCabe analysis in effect produces a “decision tree” with all “state” at the transitions from node to node. This can be used for a limited form of rules based reasoning which can highlight where vulnerabilities are or can be formed.
This in turn highlights not just “vulnerability instants” but “vulnerability classes”.
All pre 1990’s OS’s or those built on them have a very limited form of security model and this was due in the main to “resource limitations”.
Thus due to “legacy effects” this carries forwards into the present day unless specific preventions or mitigations are put in place.
Ideally the vulnerabilities would be removed and cease to be an issue. But this is not always possible due to “required functionality for legacy programs”. Thus mitigations are imposed as a protective layer between the vulnerability and any program that might exercise it, this is what various “role-based” and similar security can do. Think on how SElinux and sinilar provides finer control than the underlying OS does. They do not remove the vulnerability from the OS but “IF” –and it’s a very big if– they are properly configured for all uses prevent an attacker from accessing the vulnerability.
Oddly, the use of this type of analysis will enable the likes of automatic generation of SElinux and similar mitigating rules.
[1] Back in the mid 1970’s Thomas J. McCabe developed Cyclomatic complexity as a way to analyse a programmes code complexity as a software metric.
https://www.geeksforgeeks.org/dsa/cyclomatic-complexity/
It does this by what some have described as “walking the flow chart”. It uses this to produce a simple quantitative measure of the number of linearly independent paths through the program code. Whilst originally it was done at the source code level, it can actually work as an analysis tool as far down the computing stack as you like. Over simply it’s a two stage process,
1, Walk the code to produce the programs “control-flow graph”.
2, Count the nodes and directed edges of the graph to produce the quantitative measure.
The hard part is producing the “control-flow graph” from the program code, but once you have this “flow chart” you can use it for all sorts of other analysis. However often the best first step is to turn the “control-flow graph” into a “decision-tree” such that all paths can be found (this may not always work if control-flow can loop indefinitely).
One such use for the decision-tree is for a very limited form of reasoning based on “walking the tree” to find all possibilities or importantly just some desired possibilities. That is to use it in LLM type systems by getting “agents to walk the tree” –graph– and find all the “state” at each stage and how it is used or ignored and what consequences this has.
https://www.geeksforgeeks.org/machine-learning/decision-tree-introduction-example/
Subscribe to comments on this entry
Sidebar photo of Bruce Schneier by Joe MacInnis.
J van Prooijen • May 23, 2026 5:03 AM
I read the story about kids faking age verification. Later I realised there are some general rules behind that nobody mentioned. So I do.
If you train AI with images of honest people, it will analyse honest people well and dishonest people not well.
Access management is about keeping dishonest people out, not only about letting the good ones in.
So if you use AI for age verification for access management it should be trained with images of dishonest people. And if not done, it will fail.
In this case AI should have been trained with kids using the tricks from the theater and cinema business.