Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> With advances in computing power and unsupervised machine learning, correlating a person with the person's surroundings is now largely or fully automated – what used to take forensic experts weeks, months or years to hunt down a person by poring over indirect clues in photographs or CCTV camera footage, is now almost instantenous.

I don't know of any system capable of doing what you describe. But if such a system existed it could just use publicly available videos from Facebook and Instagram. It wouldn't need its own social media platform to feed it data.



Such a system would be built on orders of the state and it would not be publicly or widely advertised, esp. in mainland China. Russia, for instance, has built SORM (deep packet introspection of the ISP traffic at the nationwide scale) – to monitor all internet activities of its citizens (on demand, not constantly), and its technical architecture is not widely known.

Having a direct feed of videos being uploaded from the user's device is also advantageous as the higher resolution of the original video will provide more details before it is recomompressed for long term storage in a smaller resolution. Most importantly, Instagram and Twitter won't allow uncontrolled access to the content they host whereas a Chinese company simply does not have a choice and has to serve the content up on its state demands.

Until proven otherwise, it is safer to project that such a system exists.


Deep packet introspection is a completely different problem from the kind of facial recognition of people in video backgrounds that you envisage. Videos uploaded from cell phones are already MPEG-4 compressed and don't need re-encoding to be served. While it is true that social media companies don't allow unfettered access to their data, server farms for scraping web sites are a dime a dozen. I don't agree that it is safe to assume that systems there is no evidence for exists. It's one step removed from Chinese mind control viruses.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: