Sunday started with an invited talk by Véronique Cortier (INRIA Nancy) on voting, covering much of the field I've been working on in the past years. The part most people remembered in the evening was the example of how Texan farmers apparently use a physical threshold padlock scheme (though even Véronique couldn't tell us what exactly they're protecting with it). The whole talk was worth getting up early in the morning.
We had two sessions on cryptography in the afternoon but the session that I enjoyed most was the one on usable security in the morning; I wouldn't mind if having a mandatory usable security session in every conference to get the word out that humans behave in certain, scientifically understood ways and we shouldn't forget that when designing security protocols (insert standard rant about SSL certificate warnings here).
One interesting finding I took away from the talks on usable security is that if you ask people what security problems can occur for a specific task such as "logging in to a website", they will cite things like weak passwords, keyloggers etc. but as soon as you put the whole thing in a broader context of a primary task such as "an app for ride-sharing", people will list mostly task-specific risks such as "getting kidnapped" and not mention the IT security angle at all.
Also, risks linked to a primary task tend to elicit statements in a personal language "I'm afraid that I could get kidnapped" whereas IT risks were, if at all, described in an abstract and impersonal language "There could be fraud".
There is a lot that I could say about the banquet in the evening, but all I'll say here is that the format was unexpected yet enjoyable and the food was good, which is what really matters. We were also treated to an open-air thunderstorm that beats any firework show I've seen so far.