Hacker News Reader

>Someone that does not want to be tracked ... will always be able to do that.

I'm sure the Russians doing the Sailsbury attacks didn't want to be tracked by cameras but they were. If top Russian agents can be then probably most other people can be too. https://www.independent.co.uk/news/uk/crime/salisbury-novich...

>the ultimate aim of these ubiquitous surveillance systems isn't really reducing crime, but social engineering

In the UK we've had fairly ubiquitous surveillance systems for a while. In practice the uses have been reducing crime and doing people for minor traffic offences, including me - they uploaded a nice video of me turning right in spite of a no right turn sign, rather to my surprise. I've noticed no social engineering - maybe that's to come?

Technology to produce fairly convincing fake videos is already mainstream, and will only get better. For how much longer will video "evidence" be considered proof of anything?

I guess most evidence can be tampered. A lot of CCTV footage is no good as evidence because it's too fuzzy but if you can figure who did it then you can get other kinds of evidence, stolen goods etc.

1. The correlation between surveillance increase and crime decrease does not prove causation. 2. As noted in the article video survelliance is not the same as facial recognition. The latter is far more powerful.

> There might be a race between the recognition software and the circumvention technology

I wanna see someone successfully circumvent scent-based tracking. A dog has a sense of smell a million times more sensitive than a human and can track people over miles, even when a trail is days old. Imagine a network of 100,000 passive scent trackers as sensitive as dog noses, hooked up to the same centralized AI and recognition system.

With a single high-confidence hit, they can localize you. If you're on foot, you're not getting very far in any reasonable time frame.

We don't stand a chance. Sitting ducks, as it were.

You just better hope to God (or the AI, apparently) you're never a target of this system, inadvertently or intentionally, because that race is over, and humans lost.

>A dog has a sense of smell a million times more sensitive than a human and can track people over miles, even when a trail is days old.

Dogs don't work very well in urban areas and there's plenty of tricks to throw dogs off your trail.

Even in rural areas, getting in a car usually means the end of where a dog can track someone.

I did not mean literally using dogs, but passive sensors installed, e.g. in lampposts.

I'd like to see your tricks for dealing with 100,000 such sensors, one on every street corner.

(also: https://www.canidae.com/blog/2015/10/can-you-throw-off-a-tra...

"In fact, Adam found that it was basically impossible to throw off the scent of a Bloodhound. He crossed a 100 foot wide, waist high river. He changed clothes, poured cologne and coffee on himself, threw smelly fish along the trail and used scent enzyme erasers to try to eliminate his scent.")

May the odds be ever in your favor!

Does such a scent sensor actually exist outside the nose of some animals?

Maybe, there are the electronic noses: https://en.wikipedia.org/wiki/Electronic_nose But I'm not so sure how sensitive they are.

You can also classically condition a bee with food rewards to extend its mouthparts whenever it detects a specific odor. You then put the plastic bee cartridge into a device such that when the mouthparts extend, an LED illuminates.

Cheaper than dog noses, but requires a sample of the target odor and some training time. It is conceivable that the "most wanted" could each have thousands of bees trained on their scents, and then those swarms could be divided up onto plastic cassettes, which would each be loaded onto a lamppost. If the targets go near any lamppost, their bee in it tattles its location in real time. The disadvantage is, of course, that you have to go around swapping out old and dead bees for live, fresh ones all the time.

An electronic nose can just report its odor profile, and a central processing server can attempt to match and track targets. All that needs is a bit more bandwidth. But it is likely slower and lacking specificity. You could likely smell explosives or bomb making residues or gunshots or narcotics, but not specifically track Suspect X across the city. That kind of precision would require that people line up at the lamppost to be swabbed, and their swab vaporized and run through a gas chromatograph for a minute or two, which I don't see happening, even in authoritarian Singapore.

> I'd like to see your tricks for dealing with 100,000 such sensors, one on every street corner.

Misdirection and decoys. How much use are those sensors if they're all detecting you at the same time?

If they become prevalent enough, I could imagine the development of "scent bombs" that would spew out so much of your odor that the sensors for miles around would all detect "you."

Change up your cologne every few days.

It may happen that the approaches usable against automatic facial detectors would be pretty noticeable for humans around [1], thus rendering them pretty hard to use in practice, and also a target of anti-facial-recognition-paint recognition algorithms.

[1] https://www.google.com/search?tbm=isch&q=dazzle+paint+face+f...

"Someone that does not want to be tracked by facial or gait recognition, either because he's a criminal or a political activist"

Or just someone that doesnt want to be tracked by everyone..

> Someone that does not want to be tracked by facial or gait recognition, either because he's a criminal or a political activist, will always be able to do that.

But if the system is not able to assign a face to a specific identity, it might immediately mark them as "suspicious". You'd have to bypass the system by really putting on the appearance of a face within the database.

Yes, someone with unknown identity, using a burner phone (or no phone) and only spending cash is going to stand out.

State actors or suitably competent criminals may be able to use fake personas though.

> Someone that does not want to be tracked by facial or gait recognition, either because he's a criminal or a political activist, will always be able to do that

I wouldn't be so sure. How well can you control your stance or manner of walking for example?

Someone (human) trained to differentiate authors of handwritten text can often still do so even if an author is actively trying to write differently.

I think ML is great for comparing these highly subtle traits.

a wheelchair seems like an easy bypass, albeit one with very specific use.

I can see all subversives having wheelchairs in not-so-distant future.

Heelys! Never underestimate the power of fads and fashion

Crutches, gait-modifying shoes, roller-skates

There are lots of options

Admittedly haven't explored this myself, but I would expect some insert in my shoes to significantly alter my gait. When I received orthopedic soles, these changed my whole stance and gait. Heck, if there's some twist in my socks, my stance and gait are changed as well. Curious to learn more how well gait recognition algorithms can remove these effects from their prediction.

> how to control your stance or manner

Quite "easy". Put a pebble in your shoe, or some scotch tape on your skin. You'll behave differently

> Someone (human) trained to differentiate authors of handwritten text can often still do so even if an author is actively trying to write differently.

That was never proven. Please cite a (good) source or I'll take it to be false

Time to bring up this educational video: https://youtu.be/JASUsVY5YJ8?t=280

> Please cite a source.

I can't, though I think you could argue about it deductively.

> That was never proven.

I guess you can't provide a source either :)

> Put a pebble in your shoe, or some scotch tape on your skin.

If someone you know well did that, would you be able to differentiate his limp from someone else's? I am sure you could. I think it's intuitive that our "limp" betrays features of our normal behaviour.

Genuine question, what is a terrorist ? It seems to be a very fluid definition these days. Anybody can be deemed a terrorist by any establishment. Also, this kind of thinking reinforces the whole "Well, I don't mind being watched by government agencies because I have nothing to hide" reaction, and it's quite frankly worriying.

I'm open to discuss this further though

The generally accepted definitions center around the use of violence to incite political change or polticial pressure by nonstate actors. Word in varies but that's the general gist.

I know that the word gets thrown around a bunch, but it's a pretty simple idea. There is some controversy about if a state can be a terrorist, but typically we just call it state sponsored terrorism and move on.

2000 definition: someone who blows things up for political reasons.

2018 definition: someone who makes angry posts on social media.

I'm exaggerating a bit, but it's headed in that direction. If you want to see something truly worrying, look up the handful of idiots trying to promote the notion of "stochastic terrorism". The gist is that non-illegal statements which somehow get interpreted by lone nuts as incitement to violence should be categorized as a form of terrorism. Basically, they're hoping to criminalize angry rhetoric that stops well short of calls to violence. The good news is that they've been flogging this idea for a long time (pre-2016) and it still isn't really gaining ground.

>Someone that does not want to be tracked by facial or gait recognition, either because he's a criminal or a political activist, will always be able to do that.

Tons of political activists (and of course tons of criminals too) have been caught by way simpler and easier to trick methods.

> circumvention technology

Like hats, glasses and beards?

> Someone that does not want to be tracked by facial or gait recognition, either because he's a criminal or a political activist, will always be able to do that. There might be a race between the recognition software and the circumvention technology, but a dedicated individual will always be able to avoid that.

One issue is disseminating the knowledge needed to defeat such systems. I'm sure the CIA and KGB have manuals and training programs to teach those skills [1], but I'm not sure if the information is publicly available in a readily accessible form (e.g. not spread in small pieces among 100 books).

[1] https://www.youtube.com/watch?v=JASUsVY5YJ8