Schneier on Security

Nick P • March 21, 2014 2:13 PM

I love that Ross Anderson liveblogs these as it saves me both a trip and reading papers that don’t interest me. Here’s my brief thoughts on some of the presentations.

Ansari’s digital check forgery paper is good work if only to provide evidence of how weak Check 21 is compared to old methods. Might be useful in DC debates later on.

Murdoch’s security protocols and evidence work focuses on something that doesn’t get as much attention. That he improves EMV in a practical way is a good thing.

Smyth’s coversion of e-voting schemes to auction schemes is interesting. Not sure what I think of it at the moment. I do think any work that analyses or improves on secure voting systems is a good thing if only that it benefits secure voting efforts. Civitas is a good one with plenty potential for improvement, as paper shows.

Roesner’s paper on Snapchat is very relieving to this pessimist: it shows people don’t trust it enough to do risky stuff. That might be a success in security awareness. An uncommon, awesome result with the social networking generation imho.

Smith’s metadata awareness work was nice for both helping people understand metadata problem and helping us understand what people think/want regarding it.

Durumeric’s Outsmarting Proctors with Smart Watches applied a Pebble smart watch and covert voting algorithm to cheat multiple choice tests. Brilliant and a sign of things to come.

Androulaki’s Secure Data Deduplication Scheme for Cloud Storage explores a field that has many decent schemes in it already. However, her method improves against content-guessing attacks with a clever trick. It needs more analysis for its security & to see if it can be integrated with other efforts.

Maurice’s virtualized GPU paper showed that several processes sharing a GPU without memory cleaning can result in data leaks. No kidding. Orange Book era covert channel analysis showed that any resource shared by several processes can create storage or timing channels. This applied to clocks, storage locations, devices, memory, cache, etc. This should be assumed by future research so we don’t waste more time saying sharing this specific device leaks data. Assume it will unless prevented by design. End of story.

Tibouchi and Wustrow showed ECC still has plenty problems. I was right to recommend people stay with RSA until ECC gets more analysis. My recent recommendation for efficiency and timing channel resistance was Bernstein’s version of ECC. I’d like to see such analysis of it to see if design or implementation has weaknesses.

Hileman’s alternative currencies paper looks interesting. I’ll probably read it.

Dimitrienko devises a way to use Bitcoin in offline cases without double spending. It’s interesting seeing Bitcoin easily solve the types of problems that plagued ATM’s in the past.

Ian Goldberg, ever the hacker, creates a trick for semantically secure cloud encryption. It’s one of those ideas that’s so simple it might just work haha.

Dingledine shows us Tor is at 800,000 users and 30Gb of traffic. A much larger crowd to get lost in than in the past. The problem is legitimate sites are blocking Tor users due to problems they can cause. I’m not sure his solution will work.

Brenner talks about which certs to remove. This comment stands out: “With 1500 trusted CAs from 650 organisations, it’s weakest-link security on a global scale.” That there are certs that haven’t signed anything is also odd. I’m for tracking both how these CA’s are used and which lead to attacks. We can penalize those pushing poor security by removing them from major browsers or treating them as a risky site. The whole model sucks, though, in my opinion.

Hopper shows the Tor hidden service model has major architectural problems when botnets are concerned. Part of it is the key exchanges. If there’s a way to reduce them or make them faster, might reduce the effect of this problem. One reason I’m a fan of designs that get rid of public key encryption, although there’s good reasons to have it here.

Vasek makes an argument for case control studies. I think the paper has value in providing more empirical evidence that popularity is the main cause of vulnerabilities in typical software. More people use it, more it’s attacked. Supports my security through diversity principle as it’s one of only proven methods of reducing this risk.

Dmitrienko shows most OTP implementations are trash. We’ve discussed some here and predicted this. Still have to shake my head at how pervasive it is. Can’t these companies implementing security techniques get a reputable security engineer to review it? The baseline should be better. Apathy is most likely cause.

Senftleben creates a microblogging alternative to Twitter with better privacy and censorship resistance properties. This field is slowly growing. The interplay of geography, mobile and publishing gives it extra potential for out of the box thinking.

Sirer shows Bitcoin mining is way easier to game than thought. Proposes interesting fix. I wonder if all the projects utilizing Bitcoin tech, such as Bitmessage, are looking at their own designs for problems stemming from Bitcoin protocol weaknesses. If they leverage Bitcoins known properties for security, then the security might be weakened or nonexistent when known is proven false.

Spagnuolo’s Bitlodine tool profiles the blockchain to break the anonymity of quite a few users. Bitcoin is not to be trusted for anonymous transactions. It might take an exchange or another protocol built on top of it to achieve anonymous transactions. Don’t hold your breath waiting for those. Cash-based methods are still best for anonymous purchases.

Bonneau tries to solve the problem I just mentioned with mix networks. Far from solved.

At least a few practical items at the conference. That’s always good.