4 ms·
End-to-end encryption doesn't solve your problem when the attacker is your first link network provider, without out-of-band signaling. Which I believe was pare
by ethbro 8y ago
End-to-end encryption doesn't solve your problem when the attacker is your first link network provider, without out-of-band signaling.
Which I believe was parent's point. Attestation and authentication of network infrastructure (in a way that would preclude Stingray attacks) is equally important as anything you layer on top.
Otherwise, without some sort of global PKI you can consult out of band, you're just rolling the dice that the cellular station you're connecting to isn't rogue and performing a man in the middle attack.
- philsnow 8y agoIf you have to trust your first network hop, then the "ends" of your "end to end" encryption are not the actual ends. The trust boundary needs to be pushed all the way up into the application on both sides. Some sort of global PKI like what iMessage uses under the covers?
- mikeash 8y agoWithout some sort of global PKI, how do you know your data connection actually goes to your cell provider?
- ethbro 8y agoI assume the way most embedded devices guarantee security for their limited number of use cases -- pre-seed trusted root keys in hardware + validate signing against them when connecting. My point was that app-layer security magic doesn't mean much if your first network step is "connect to closest, highest power base station and implicitly trust it." There are wonderful things we can do on top of lower layer protocols, but the lower layers are pretty damn important too.
- mikeash 8y agoThat’s how, for example, TLS works too. I don’t understand what you mean when you say that app-level security doesn’t mean much if your first network step is compromised. TLS will protect you there, as will any other decently implemented end-to-end encrypted protocol.
- ethbro 8y agoTLS has the benefit of a world-wide PKI built around it to attest to authentication. I was trying to note the risks to PKI-less / web of trust participants, and should have called that out explicitly. Mostly because I'd much rather live in an encrypted world where WoT is the dominant mode, vs corporate-controlled PKI. And the biggest risks there seems like the step no one can get around in ordinary use.