This week I’m attending a Dagstuhl Seminar in Wadern, Germany. This well-known seminar series gathers computer scientists from around the world to discuss and brainstorm on a significant research challenge related to computer science.
In this particular seminar we’re discussing on the Foundations of Secure Scaling. It’s a topic that was proposed by a few colleagues - Lejla Batina from Radboud University, Swarup Bhunia from the University of Florida, and Jean Pierre Seifert from Technical University of Berlin, and myself.
What is Secure Scaling, and why do we have a seminar about it?
To explain secure scaling, we should start with scaling by itself. The effects of scaling on digital designs are well known. Moore’s Law has fuelled a drastic increase of the logic density in digital designs. This, in turn, enabled complex digital architectures such as System-on-Chip (SoC). The SoC technology has propelled ever more complex integrated applications, to the point where we can now make casual phone calls on billion-transistor computing devices (smartphones). So technological scaling has a clear effect on the capabilities at the highest abstraction level of design.
But for secure design, the effects of scaling are far more complicated and ambiguous. The security of a chip does not depend on performance factors such as power or speed. The security of a chip depends on its ability to support cryptographic algorithms and their associated security policies, such as how to handle secret keys, and how to provide guarantees on correct execution. In other words, faster chips are not more secure; they are just faster. To build secure chips, we need a better insight into the needs for safety and security, and into the threats that seek to eliminate it.
The seminar hosted 26 researchers from all continents, each of them a member from one of several communities. There were cryptographers, system integrators, and technology experts. Above all, it was a highly motivated group of individuals willing to think and talk about hard problems.
Scaling has a different meaning depending on who you’re talking to, a cryptographer, a system integrator or a technology expert. In the seminar we had talks on the often fascinating mechanisms of scaling in each of these communities. For example, cryptographers are faced with a future Internet full of Things that may have significant implications into safety and privacy. System integrators are faced with the daunting task of composing components that may or may not be trusted, into safe and secure systems. Technology experts deal with sophisticated tampering mechanisms as well as complex implementation effects that have repercussions to the highest abstraction levels of design (think about the recent progress in adapting side-channel analysis and fault-injection analysis to the cloud, for example).
So given the challenges of secure scaling, what are these three communities to do? There are no easy answers to difficult problems. This Dagstuhl workshop is not looking for an easy solution, but rather for a better problem statement. As researchers, we should not feel compelled into thinking that our own short-term solutions are a hammer that can nail every problem. A workshop such as this one, looking into the different aspects of a common long term challenge, is an opportunity to improve not only our research but also ourselves.