Facebook to scan users' photos for paedo pics
Discussion
Telegraph said:
Facebook to trawl photos for child abuse
Facebook will introduce technology to automatically trawl photos posted by users for child abuse in an attempt to drive paedophiles away from its service.
PhotoDNA, a system built by Microsoft, will mean Facebook will not have to rely on complaints to police the billions of images uploaded by its more than 500 million users.
The system is run by the National Center for Missing and Exploited Children (NCMEC), based in Virginia. The organisation has collected 10,000 images of child abuse from law enforcement authorities to serve as the basis of PhotoDNA, and has millions more on file.
Computers divide the images into smaller sections and calculate a hash, or digital fingerprint, for each. The hashes are shared with web services such as Facebook, which can then automatically compare them to images uploaded by users.
According to Bill Harmon, a lawyer in Microsoft’s digital crimes unit, PhotoDNA detects child pornography with “zero false positives”.
“Some images become ‘popular’ and are used time and time again – making good targets for the PhotoDNA program,” he said in a post on the Microsoft blog.
(Full article here: http://www.telegraph.co.uk/technology/facebook/852...)Facebook will introduce technology to automatically trawl photos posted by users for child abuse in an attempt to drive paedophiles away from its service.
PhotoDNA, a system built by Microsoft, will mean Facebook will not have to rely on complaints to police the billions of images uploaded by its more than 500 million users.
The system is run by the National Center for Missing and Exploited Children (NCMEC), based in Virginia. The organisation has collected 10,000 images of child abuse from law enforcement authorities to serve as the basis of PhotoDNA, and has millions more on file.
Computers divide the images into smaller sections and calculate a hash, or digital fingerprint, for each. The hashes are shared with web services such as Facebook, which can then automatically compare them to images uploaded by users.
According to Bill Harmon, a lawyer in Microsoft’s digital crimes unit, PhotoDNA detects child pornography with “zero false positives”.
“Some images become ‘popular’ and are used time and time again – making good targets for the PhotoDNA program,” he said in a post on the Microsoft blog.
Am I being incredibly naive here? Because to my mind nobody with anything to hide (like the aforementioned paedos) are going to actually upload illegals pics to a Facebook photo album, are they? Or maybe I'm wrong and they do, and this kind of scanning will be beneficial.
ChiChoAndy said:
"Zero false positives"? I fail to see how something like that can give zero false positives?
Indeed. There will be hash collisions - they're unavoidable. I just hope that it flags the pic up for review by a human rather than just deleteing it and automatically suspending your account.Knowing Facebook though, it'll be the latter.
JonRB said:
Because to my mind nobody with anything to hide (like the aforementioned paedos) are going to actually upload illegals pics to a Facebook photo album, are they? Or maybe I'm wrong and they do, and this kind of scanning will be beneficial.
I dunno,Stranger things happen. Take this guy for example;
http://www.10news.com/news/27931741/detail.html
He admitted to Child Abuse whilst applying for a job.
But not just any job......he was applying to be a police officer
Back on topic;
Though the intention behind this is good, it's seems to me like it's just more intrusion into people's privacy.
IIRC Microsoft is a big investor in Facebook, hence the bing tie in etc. Microsoft are also behind this system so looks like a marketing ploy rather then anything that's actually going to be much use.
Saying that, these online paedo's aren't always as discrete as you might expect as a lot of the operations against them prove.
Saying that, these online paedo's aren't always as discrete as you might expect as a lot of the operations against them prove.
Shuvi Tupya said:
ChiChoAndy said:
"Zero false positives"? I fail to see how something like that can give zero false positives?
Agreed, there must be pictures on facebook of naked kids that are perfectly innocent and legal to own, this software must flag those up too?Not that it matters i suppose!
I think that one (rare, but possible) scenario would be where a parent has uploaded an innocent image which the paedo's like for some reason and it gets on the database. The scanning software would then generate a 'hit' when it came across the 'original' which had been posted innocently.
Emeye said:
Not got chance to read the article until later, but I suspect the software can't tell if the picture is dodgy, but checks to see if any of the uploaded pics are ones which are already identified as being in circulation in the kiddie fiddling world.
You're correct. A hash (a sort of fingerprint) is calculated for the pic and compared against a lookup list of hashes of known dodgy pics. However, hashs are not unique - it's statistically possible for two different images to generate the same hash. And it is also possible for two pictures that are identical to the human eye to have different hashes if they have been altered in some way from each other.
mrmr96 said:
I think that one (rare, but possible) scenario would be where a parent has uploaded an innocent image which the paedo's like for some reason and it gets on the database. The scanning software would then generate a 'hit' when it came across the 'original' which had been posted innocently.
That's a plausible scenario, and one that hadn't occurred to me either. Edited by JonRB on Friday 20th May 15:38
Shuvi Tupya said:
Think what else this software could be used for though...
They could stop you from posting any picture they don't want seen.
Not really. As has been mentioned above, you can have two pictures which look the same but have different hashes, if one has been digitally edited in some subtle way. (e.g. change the resolution, or crop it, or add a watermark etc)They could stop you from posting any picture they don't want seen.
It seems they do Jon
http://www.bbc.co.uk/news/uk-scotland-glasgow-west...
A man who sexually abused a two-year-old girl was caught after he posted images of the attack on Facebook.
http://www.bbc.co.uk/news/uk-scotland-glasgow-west...
A man who sexually abused a two-year-old girl was caught after he posted images of the attack on Facebook.
Penny-lope said:
It seems they do Jon
http://www.bbc.co.uk/news/uk-scotland-glasgow-west...
A man who sexually abused a two-year-old girl was caught after he posted images of the attack on Facebook.
So it does scan for new images, and interprets if it is child abuse? http://www.bbc.co.uk/news/uk-scotland-glasgow-west...
A man who sexually abused a two-year-old girl was caught after he posted images of the attack on Facebook.
mrmr96 said:
No, I don't think that it can work out what's 'happening' in a photo. It's basically looking each photo up against a database of KNOWN images to see if the picture is on the database.
But that fellow was done 2 days after an attack, and posting the photos on facebook? Ah... Missread the article.
ChiChoAndy said:
So it does scan for new images, and interprets if it is child abuse?
No...the bloke had 7000+ images of abuse.They wern't all him and the kiddy some must have been "classics" (if you pardon the expression)
It also doesn't mention the software anywhere in that article.
It could have just as easily been a tip off.
I have seen a few different people post up pictures of their own kids on FB where they are naked. All innocent stuff (baby in the bath, toddler on beach, etc) but it did strike me as odd. No way would I post up stuff like that of my own kids on a semi-public forum. Would this software detect such things? Could they get into trouble? (i.e. should I warn them?)
Splats said:
I have seen a few different people post up pictures of their own kids on FB where they are naked. All innocent stuff (baby in the bath, toddler on beach, etc) but it did strike me as odd. No way would I post up stuff like that of my own kids on a semi-public forum. Would this software detect such things? Could they get into trouble? (i.e. should I warn them?)
As the other posters, and the article explains - it doesn't detect content of pictures, it reads the 'tag' of sorts, and compares them to an existing database of such pictures. Unless your friends are posting pictures which have the same 'tags' as images from this database, there will be no problem whatsoever.Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff