site banner

Small-Scale Question Sunday for December 18, 2022

"Someone has to and no one else will."

7
Jump in the discussion.

No email address required.

How will future security cameras defeat the inevitable "deepfake defense"

Even if fakes become indistinguishable from real footage, it'll still remain valuable. After all, there are plenty of forms of evidence we accept today that are even easier to fake.

Eg. we accept witness testimony, even though just lying is even easier than photoshopping. So I think we'll just see the same standards applied: we'll trust footage on the basis of things like whether you have a motive to frame those particular people etc. So it's still going to be stronger evidence than, say, just reporting that you saw them, since doctoring footage would require actual malice on your part, while you could have been mistaken if it was just your recollection. And it should also be noted that just being indistinguishable from a fake "from the pixels" doesn't make it impossible to distinguish in general: like all lies, you're opening up yourself to contradictions if there are other reliable pieces of evidence to compare with (eg. if the kids have an alibi, or if your footage differs from your neighbours in terms of things like expected shadows / ambient light etc for the same time, you could be caught out). Even if no such evidence happens to exist, you can't reliably know that in advance, and so the fact that you're implicitly opening yourself up to prosecution in turn should it be wrong makes your footage more credible.

There are also probably ways to improve this further should such claims become more prevalent. Eg. cameras handled by third party companies that archive the realtime footage and provide a documented and consistent chain of custody for the evidence: faking footage in such a system would require a lot more technical knowledge.