Certainly targeting low hanging fruit where there is a strong case for network/law enforcement collaboration, it would make a meaningful difference and it is relatively easy to do, makes sense.
The question is whether any of these really exist.
It may be that social networks like Facebook are a reasonable place to start in that Facebook do have access to the communications content at rest and already have technology in place for authorised parties to access it.
I doubt this would help catch the organised criminals though — maybe a few not-so-bright individuals would get caught in this net but that’s about it.
So it is wrong for us to swallow the false hope that this will do anything except displace bad actors onto other technologies, that’s all that will happen.
After I wrote the original medium piece, I came across this document called “Keys Under Doormats” http://dspace.mit.edu/handle/1721.1/97690#files-area which is an old MIT taskforce from the 90s who have reconvened in light of the new pressure to provide government access to encryption.
Their document highlights 3 key areas of concern:
- Breaks current best practice of using forward secrecy — this is an approach that means that keys are deleted after each use meaning that prior or later comms are not compromised (your point about data in motion)
- Substantially increases system complexity — because offering keys in escrow is never simple this increases complexity, in many experts minds the number one enemy of security.
- Creates concentrated targets (points of failure) — if there was a keyholder for everyone’s encryption then that keyholder would be subject to more attack than anyone else.
The debate however is not going away anytime soon. I notice that Indonesia banned Telegram yesterday. Whether this moves everyone onto Signal instead is not yet clear.
