Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

washingtonpost.com

Post reply on HN
Print view View original

141–150 of 154 posts

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#141

Post by tomc1985 »

pavel_lishin wrote: post #116
kardos wrote: post #99

Earlier quoted context omitted.

That sounds a lot like security by obscurity, one blog post later and the world can distinguish the two. Cvdazzle is unlikey to be the answer, it's just today's CAPTCHA to slow down face recognition, tomorrow's recognition will cut through that We need a more radical solution to devalue our private information. Consider by analogy cell phones: IMEI blacklists (in theory) fundamentally devalue stealing phones. What ca…

> if holding personal information was a big liability in the event of a leak, it would be prudent to keep as little as possible. But this kills the social network, and the cloud photo app, and a lot of other businesses. edit: nevermind, I think I misunderstood the suggestion. kardos's reply makes a lot of sense.

Good!

Or perhaps it would result in only the companies that need that data actually holding it, and doing so with care. I would love to see a big nasty liability poison pill attached to PII

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#142

Post by dingaling »

praeivis wrote: post #2

But only if you published your photos in public. If you dont spread your photos around the web you still safe in crowd.

I have a colleague / acquaintance who worked laboriously to prevent photos of himself being put online. He avoided all social media, tried to duck out of group photos and used pseudodyms for fora. Searching for him only returned photos of other people sharing his name.

Then one day last year he attended an open-day about retraining IT professionals as teachers. He didn't read the small print closely enough and photos taken at the event were published and studiously tagged by the event organisers. There he was, bubbling up on Google Images.

Subsequently he contacted the organisers and they did remove the tagging, but it was purely by chance that he had discovered it.

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#143

Post by moheeb »

dclowd9901 wrote: post #87

> The algorithm comes from Moscow-based NTechLab, which recently boasted it bested Google’s recognition software at MegaFace, a technical challenge hosted by the University of Washington. It's a bit baffling to me why University of Washington was trying to accelerate this capability. I'm all for pure academia, but this is just fingerprinting someone's face. The real tech is in being able to map a face from a 2D image…

Fingerprinting turned out to be one of the biggest crime-fighting tools of the 20th century. Does that debaffle it a bit?

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#144

Post by sbov »

imgabe wrote: post #140
sbov wrote: post #136

Earlier quoted context omitted.

This distinction seems relevant only if you target one person, which seems pretty rare. In reality you're going to use this metric on a pool of people, in which case 65% of them would become an alcoholic in 5 years.

It becomes a problem when it's used for profiling, such as "Don't hire that person, there's a 65% chance they'll be an alcoholic" or "send a drone to kill that person, there's an 80% chance they're a terrorist"

I agree that it's a problem, but your distinction still seems generally pointless. If you have a company policy to not hire alcoholics and you reject 1,000 people based upon this metric then 65% would have been an alcoholic.

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#145

Post by nitrogen »

dclowd9901 wrote: post #91
kardos wrote: post #21

Earlier quoted context omitted.

That's a really hard battle to fight, even for those of us who understand the full implications. The entire population is not going to drop facebook overnight, it does deliver some sort of value to them despite it's heavy privacy price. I'm beginning to believe that the answer is to simply devalue all that information. We can do it in some contexts on the technical side, for example, rotating passwords devalues stole…

You don't need to drop Facebook -- just turn privacy settings so that your face can't be identified publicly or beyond 1 degree of connection. People should know how to do at least that .

Your main profile picture is always public, as is your cover photo, so you will have to use a picture that doesn't identify you, which may be confusing to your friends.

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#146

Post by nitrogen »

qaq wrote: post #119
dangrossman wrote: post #107

Earlier quoted context omitted.

Employers routinely subject applicants to personality tests, IQ tests, fitness tests for jobs that have nothing to do with manual labor (essentially age/gender discrimination), background checks, credit checks even when the job doesn't involve handling money, and yes, looking them up on social media. Do people really want to work for these places? Yes, because they need to pay their rent, and don't have the nestegg o…

This holds true for 3d world countries in US for 90% of people the only reason to be in this position is one's own choices.

The sense of free will is an illusion of physics and consciousness; choices are made by quasi-deterministic electrochemical processes before our conscious minds are ever aware of them.

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#147

Post by AJRF »

miander2 wrote: post #126
AJRF wrote: post #37

Earlier quoted context omitted.

Well you can't do that. You can only tag someone with an account.

You can tag anyone on Facebook even if they don't have an account. When you start typing their name a drop-down box of suggestions appears. Hit the escape key and you can continue typing their name and tag them. Anyone with your photo can associate it with your name on Facebook even if you don't use it.

Ah just check and yes you are right, apologies. Do people actually do this? Pretty ignorant if they do!

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#148

Post by pradn »

aub3bhat wrote: post #49
wangii wrote: post #30

>> “With this algorithm, you can search through a billion photographs in less than a second from a normal computer,” With Intel Core i7 5960X (238,310 MIPS at 3.0 GHz), it only takes 238 instructions for a photo?

The Google FaceNet and other similar system typically perform an approximate nearest neighbor search in a ~128 dimensional embedding of faces. Here is the paper for the reference: http://arxiv.org/pdf/1503.03832v3.pdf Also 70% accuracy is like really really bad. Here are results from MegaFace Challenge, which involved identifying face in presence of Million distractors: http://megaface.cs.washington.edu/results/fgnet…

The accuracy is not 70%. The percentage of times it is able to match a face to a social media profile, given the person has a profile, is 70%. You can't compare apples-to-apples with the results in Google's FaceNet.

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#149

Post by qaq »

nitrogen wrote: post #146
qaq wrote: post #119

Earlier quoted context omitted.

This holds true for 3d world countries in US for 90% of people the only reason to be in this position is one's own choices.

The sense of free will is an illusion of physics and consciousness; choices are made by quasi-deterministic electrochemical processes before our conscious minds are ever aware of them.

if we view things at this depth then this whole discussion is pointless :)

Re: Russia’s new FindFace app identifies strangers in a crowd with 70% accuracy

#150

Post by allemagne »

aub3bhat wrote: post #83
allemagne wrote: post #65

Earlier quoted context omitted.

>Why should corporations and other people be denied the same access & privilege. Should we be offering that access and privilege to governments? Assuming that we should, your argument seems to be that it's "unfair" not to offer the same to businesses? Bizarre. Perhaps you can clarify. >The reality is that most people correctly understand the trade-off between privacy and utility. I would argue against that. However,…

>> Should we be offering that access and privilege to governments? They already have it. >> your argument seems to be that it's "unfair" not to offer the same to businesses? Bizarre. In Sorrell vs IMS Health (2011) Supreme Court ruled 6-3 in favor IMS Health, that Commercial Speech was protected under First Amendment. This case is particularly relevant since IMS Health counsel successfully argued that Government coul…

Whether the government already has data is immaterial when the question is whether they "should" have it from a moral or philosophical standpoint, which is the context I'm assuming we're working within. The legality of their having it certainly has questionable constitutionality.

In Sorrell vs IMS Health, the issue does not seem to be whether government and private companies could use the same data to market pharmaceuticals, but whether pharmacies could sell prescription data without doctors' consent. "Fairness" to companies using the same data as government appears to have nothing to do with it. You missed a period at the end of your link, so I added it in mine below.

If we ignore that and assume your interpretation is correct, your conclusion that private individuals and organizations are privy to the exact same personal information that all government organizations may have is definitely not the conclusion that outcome would point towards and is still, indeed, a bizarre belief.

As to your final argument, who is talking about suing companies? If you're responding directly to something I said, the logic to connect it to your response is unclear at best.

https://en.wikipedia.org/wiki/Sorrell_v._IMS_Health_Inc.

Post reply on HN