4 ms·
The problem here is that encryption is an overloaded term. There are two different things happening when one encrypts, and both are useful independently. The
by jedberg 15y ago
The problem here is that encryption is an overloaded term. There are two different things happening when one encrypts, and both are useful independently.
The two parts are identity and obfuscation.
Obfuscation protects you from listeners who do not control the network, and identity protects you from listeners who do control the network.
Together, you know only your intended recipient is getting your message, but apart, at least you are protected from certain threats.
So even if we just had obfuscation, that would protect us from most threats. When people say "we need encryption everywhere", I think generally what they mean is "we need obfuscation everywhere" to protect from the worst threats, which are those of the uncontrolling listener.
The flip side is that the uninformed will get complacent and assume that obfuscation is giving them identity protection as well.
- Herring 15y agoIf we're going to overhaul the entire structure of the internet, it wont be because of coffee shop eavesdroppers that haven't learned how to inject traffic.
- tptacek 15y agoI think that's it in a nutshell. When we talk at the level of the design of protocols, we generally concede everything we can reasonably concede to the attacker, and then design against that well-armed attacker. Denying attackers the ability to redirect or inject traffic ties their hands behind their back. There's really no such thing as an "purely passive" attacker. Since the 1990s, probably starting with the deployment of switched ethernet, active attacks have gotten steadily easier and more popular, and passive attacks less so. The frustrating thing here is, we are so close to having this problem licked, but so many smart people are spending time concentrating on problems we don't have. What we need to do is to come up with a credible replacement for the Mozilla/Microsoft/Google-controlled CA system. It's probably no harder to solve that problem than it is to push adoption of weak unauthenticated protocols.
- simonbrown 15y agoI assume by obfuscation you mean browsers accepting self-signed certificates in some cases. To do that would require everyone to be using a browser that accepts them, nearly all don't. On the other hand, StartSSL is free and already accepted in some browsers. It seems more promising that eventually nearly all browsers in popular use will have StartCom as a root CA. Of course, there are other problems preventing small and large sites from using SSL, like the requirement of a dedicated IP address (which SNI fixed, but is again not supported in all popular browsers) and the overhead.