Blog 15-400 About Github Repository Source

Stealthy Attacker Complexity in Cyber-Physical Systems

This is the homepage for my 15-400 research project. I will be posting regular updates on my research during the S20 semester.

Brief Description

  • Who: I will be working under Eunsuk Kang, assistant professor in the Institute for Software Research at Carnegie Mellon University.
  • What: I will be studying the relative complexity between supervisors and stealthy attackers in a cyber-physical system. This will involve defining an appropriate notion of complexity that makes sense for both attackers and supervisors and then considering questions such as
    • How complex does an attacker have to be, as a function of the supervisor’s complexity, to guarantee a stealthy attack?
    • How complex does an attacker have to be, as a function of the supervisor’s complexity, to make a stealthy attack possible in some cases?
    • How complex does a supervisor have to be, as a function of the attacker’s complexity, to prevent a stealthy attack?
    • How complex does a supervisor have to be, as a function of the attacker’s complexity, to make a stealthy attack preventable in some cases?
  • So What: Nowadays, cyber-physical systems are found in all sorts of safety-critical applications, such as in automotive vehicles. Consequently, attacks on cyber-physical systems are an important concern for system designers. Upon detection of an attack, it is often a viable option for a safety-critical system to simply shut itself down, thus preventing any harm. However, such detection is by definition impossible for stealthy attacks, which makes them particularly frightening and especially important to understand. My work will enable cyber-physical system designers to obtain security guarantees on their systems. By measuring the complexity of their systems and applying the bounds I will derive, cyber-physical system designers will be able to obtain lower bounds on the complexity of any stealthy attacker, which may be useful in ruling out threats from certain classes of actors.

Research Proposal

Milestone Reports

Results