[epolanski]

I've learned from a former college colleague that got into cyber security that Israeli intelligence facial recognition is virtually error free.

It has been trained on decades of Palestinians crossing check points, some being Hamas camouflaging with beards, glasses and what not.

Also the data it's fed for third party customers is as flawless as it can be: if you ever took an international flight your biometrics are fully recorded and available to virtually every agency in the world.

If you're walking in a random mall on the other end of the world, even if you have no phone, you have covered your tracks and you're wearing a hat and glasses, etc, you are going to be recognized by the software if a camera gets even a mediocre shot at you.

Compound this with all the information people put online on their own on socials, you're gonna be tracked and recognized, whether you want it or no.

[YeGoblynQueenne]

>> I've learned from a former college colleague that got into cyber security that Israeli intelligence facial recognition is virtually error free.

What does "virtually error free" mean? There's no "error free" in facial recognition, or any other application of machine learning.

More to the point, who says all this, besides yourself in this thread? Why should anyone believe that "virtually error free" is a factual description of real technological capabilities rather than state propaganda?

[epolanski]

Those are private companies, so it's not state propaganda.

By the way, UK, South Wales especially claims an 89%+ success rate and 1 in 6'000 false positives, you can read it on UK's official website.

The company between Oosto claims 99%.

https://www.gov.uk/government/publications/police-use-of-fac...

[AuthAuth]

That is not even close to error free. 89% is really bad actually.

[epolanski]

Those are numbers claimed by the UK, the company behind it claims an order of magnitude over it with proper data (airport-level full face scans).

Even 89% isn't that bad imho, recognizing the overabundant majority of a population with random cameras that don't require the user to pose or assume specific positions is..quite something.

[AuthAuth]

Its kinda of bad when factoring in the consequences for being misidentified. Getting misidentified can cost you your life. It can cause you months of time and legal fees to disprove. It can waste police resources and erode public trust.

And I trust the UK police data far far more than the company. Every company says they are 99% accurate.

[Izkata]

It's the other number. 89% of queries result in a match, and of those the error rate is 0.02%

[ponector]

I don't know which has more lies: state propaganda or marketing materials from private company.

Claim that China will pay tariffs to the US or claim of full unsupervised autonomy driving by the end of 2016.

[joquarky]

Imagine the utility of only 99% accurate OCR.

Now apply that to something that can drastically alter someone's life

[YeGoblynQueenne]

Thanks for the link. Quoting from it:

* The National Physical Laboratory (NPL) tested the algorithm South Wales Police and the Metropolitan Police Service have been using for LFR.

* At the settings police use, the NPL found that for LFR there were no statistically significant differences in performance based on age, gender or ethnicity.

* There was an 89% chance of identifying someone on the specific watchlist of people wanted by the police, and at worst a 1 in 6,000 chance of incorrectly identifying someone on a watchlist with 10,000 images (known as a false alert). In practice, the false alert rate has been far better than this.

It looks like this is the report by the NPL:

science.police.uk/site/assets/files/3396/frt-equitability-study_mar2023.pdf

It's a big report and it'd take some time to go through it but it's clear that laboratory testing of a system deployed in the wild is not going to give accurate results, meaning the "89%" claimed is going to be significantly worse in reality. Anyway there's obvious limitations to the testing e.g. (from the report):

Large demographically balanced datasets: The testing of low error rates in a statistically significant manner requires large datasets. To achieve the required scale, the evaluation uses a supplementary reference image dataset of 178,000 face images (Filler dataset). This is an order of magnitude larger than the typical watchlist size of an operational Live Facial Recognition deployment. To avoid introducing a demographic bias due to reference dataset composition, a demographically balanced reference dataset was used, with equal numbers in each demographic category. For assessment of equitability under operational settings, the results from the large dataset are appropriately scaled to the size and composition of watchlist or reference image database of the operational deployment.

I'd say "uh-oh" to that. Unbalanced classes is a perenial source of error in evaluations. "Equal numbers in each demographic category" is an obvious source of unrealistic bias.

Anyway, I don't have the time to go through that with a fine toothed comb, but just the fact that they report a 100 False Positive Rate for "operator initiated facial recognition" is another big, hot, red flag.

Also, from the UK gov link above:

* The 10 LFR vans rolled out in August 2025 are using the same algorithm that was tested by the NPL.

There's a bit of ambiguity there. The police are using "the same algorithm" tested by the NPL, but are they using the same settings? The report uses specific settings to come up with its conclusions (e.g. a "face match" setting of 0.6 for LFR), but there's nothing to say the police stick to the same. Lots of room for manoeuvering left there, I'd say.

>> The company between Oosto claims 99%.

We can easily dismiss this just by looking at the two digits preceding the "%".

[throway23423]

I doubt whatever facial recognition trained over 6 million odd Palestinians (plus 2 million Israeli Palestinians) would trump similar offerings from competitors like Hikvision trained on data of 1.4 billion Chinese.

edit : i think their tech is overhyped. Remember the signal-chats debacle last year where the National Security Advisor was photographed using a modified client of Signal by Israeli company TeleMessage. And immediately after, TeleMessage was hacked, and it was revealed that all the chats were transmitted and stored in plain-text. They still managed to get their backup-spyware installed at the highest levels of the US government and military. It looks like they have great sales teams.

https://archive.is/0qjVI

[yes_really]

A facial recognition model trained on one genotype will behave poorly on another genotype. For detecting e.g. white and middle eastern faces, this Israeli model should perform better than the one trained on Chinese people

[lp4v4n]

I believe that most of what you said is true, but I don't think the tracking of people around the world is as efficient as your post suggests. If a single face scan were enough to track people anywhere like that, American government agencies (I'm thinking ICE, the FBI, etc.) wouldn’t have as much trouble as they do arresting people. That’s just my impression of course, maybe for some reason they choose not to use these technologies.

[peyton]

They need recall, not precision. It’s conceivably fine if you tag 100 people as long as one of them is your guy.

Also I mean you and I can recognize people we know. A surveillance camera has millions of sensors sampling every ~50 ms. It’s plausible.

[sejje]

> if you ever took an international flight your biometrics are fully recorded and available to virtually every agency in the world

Approximately what year did this start?

[epolanski]

I have no clue because my first extra EU flight has been in 2022 and I definitely got a full face scan.

[gruez]

The part I'm skeptical about is "available to virtually every agency in the world". I think every immigration checkpoint I've been to have some sort of camera setup, but the extent of data sharing is unclear. Is China sharing data with the US? Or US sharing with Canada? US with Germany? etc.

[deaux]

Except for a handful of countries like China, you can very reasonably assume all of the others' is both available to Israel as well as the US.

[nick_]

Is the Israeli intelligence facial recognition system in the room with us now?

[epolanski]

Oosto, Corsight AI.

[pdyc]

i have doubts on accuracy of face recognition. There is already nancy guthrie case going on and if it is so accurate why are suspects still not recognized?

[pear01]

You mean the the case where he came to her door dressed like death with his face almost completely covered?

Pretty extreme bar your setting. I would think most people would agree it could still have extremely (and surprisingly) potent accuracy and still fail in this case. I wouldn't expect facial recognition to work in a case when there is little to no face to work with... if that guy came dressed like that to any airport or mall he would've been detained immediately.

[riobard]

I used to think that the scenes of the TV series “Person of Interest” were exaggerated for storytelling purposes. Maybe not and it was accurate prescience.

[zjaffee]

I mean, it's objectively true that they can do this, especially when even mildly filtered down by incoming external data.

It's why you no longer need to speak with a person when reentering your home country in a lot of different places (israel being one of them, but also the EU, trusted travelers in the US through global entry, ect).

[gerash]

I'm sorry this sounds like hyperbole to say the least.

idk about their accuracy but "error free"?

Also do you understand the amount of compute and network bandwidth necessary to index and track billions of people by processing exabytes of streaming footage constantly with heavy computer vision models. who's integrating all these different camera systems to start with?

These claims sound like they come from someone who hasn't done these things in real life.

Tangentially, it seems like Israel tech scene has so many players involved in spyware/malware and surveillance.

[somenameforme]

I think these sort of claims of excessive competence are challenged by the October 7th attacks. Think about the massive amount of planning and organization that went into that attack over a period of years. There were thousands of forces engaging in some specialized and unusual strategies. Hamas even released a propaganda video more or less showing their plan with paragliders and everything. And they carried it out the day after the 50th anniversary of the Yom Kippur War. And somehow this all caught Israel completely by surprise. So either you have to go down a very dark rabbit hole, or accept that these claims of excessive competence are, at the minimum, exaggerated.

Similarly this would make things like evading law enforcement pretty much impossible, while in reality there are countless people, at least thousands, who have been photographed in relation to e.g. a crime, but never found, and never identified.

[deaux]

> So either you have to go down a very dark rabbit hole

So even after "there's a child sex trafficking island where all the elites have gone to party for decades" you're still skeptical of that claim? Knowing about Mossad operations? With Bibi on the record saying

> Anyone who wants to thwart the establishment of a Palestinian state has to support bolstering Hamas and transferring money to Hamas … This is part of our strategy – to isolate the Palestinians in Gaza from the Palestinians in the West Bank

With most of the world's spyware, including Pegasus and NSO group, having hailed from Israel?

It's not "going down a very dark rabbit hole", it's the by far most likely option and therefore your whole comment makes no sense, presuming the much less likely option.

If we're still not at the point where we stop being this naive, my god..

[delichon]

> If we're still not at the point where we stop being this naive, my god.

That is a naive statement given that 75% of the world's population identifies with an established religion, and each of those have evidence free beliefs such as virgin birth, reincarnation, the existence of hell, etc.

[Pay08]

You either have no context for that quote or have intentionally misrepresented it. Hamas started out as an arm of the Muslim Brotherhood, with both being charities at the beginning. They both became political arms a little later, with Hamas being far, far less radical than the PLO at the time. The idea behind funding Hamas was to either cause a civil war between them and the PLO or to have the less radical Hamas faction take over the PLO. Obviously, it didn't work but that doesn't mean the idea doesn't have merit and has been used elsewhere successfully (Israel didn't come up with this tactic).

[tzs]

> So even after "there's a child sex trafficking island where all the elites have gone to party for decades" you're still skeptical of that claim?

That may be a little exaggerated [1].

https://www.pbs.org/newshour/nation/ap-review-on-what-the-ep...

[somenameforme]

So the elites were flying in on the Lolita Express to an island filled with underage sex workers, to a venue with luridity abounds, hosted by a convicted pedophile, to discuss philanthropy?

Or might it be that the entire political class is filled with moral degenerates devoid of ethics, to nobody's surprise. And so consequently you are effectively having suspects investigate themselves only to conclude 'nah, nobody did anything wrong except the dead guy, but he's dead so ¯\_(ツ)_/¯ '

Even that report notes at least 4 or 5 victims stated they were abused by other men and women besides Epstein. So it's gone from #MeToo to #Only4or5. If you are ever curious what creates cynics like me...