Microsoft Tip Leads To Child Porn Arrest In Pennsylvania 353
Shades of the recent arrest based on child porn images flagged by Google in an email, mrspoonsi writes A tip-off from Microsoft has led to the arrest of a man in Pennsylvania who has been charged with receiving and sharing child abuse images. It flagged the matter after discovering that an image involving a young girl had been allegedly saved to the man's OneDrive cloud storage account. According to court documents, the man was subsequently detected trying to send two illegal pictures via one of Microsoft's live.com email accounts. Police arrested him on 31 July.
Re:Microsoft's child porn collection (Score:4, Informative)
Actually, no.
They get a big list of file hashes from the National Center for Exploited Children or something, and it's implemented as part of the file scan. All that happens is they check file hashes and if it matches, then they do more in-depth analysis (is it an image file? etc).
Which begs the question on the general stupidity since hashes are so trivially easy to change and it's extremely easy to obfuscate (just zip it up with a password).
People are lazy. Even ones who really know that what they do isn't really appreciated by the general population and really ought to try to cover their tracks... and don't.
Nope, from the TFA they process the image to derive a signature which can survive things like resizing, changing resolution etc. It's not just a simple hash.
Re:Which company is next in line? (Score:5, Informative)
NCMEC has the collection of actual illegal pictures. They have government permission to have them.
Everyone else (Microsoft, Google, Facebook, etc) just has the list of hash values. Totally legal for them to have.
This system has been public knowledge for at least 3 years. Just google NCMEC and follow the links!
And (since someone always complains) yes, the people running this know what a hash collision is. They are experts with hash functions and image processing.
Re:Which company is next in line? (Score:3, Informative)
They don't hash the raw file itself they construct a specialised hash based on the image content. It breaks the image up into chunks, analyses those chunks and generates a hash from that analysis. The intent being to make it resilient to cropping, scaling and colour changes. http://en.wikipedia.org/wiki/PhotoDNA