Identifying algorithmic vulnerabilities through simulated annealing

Real–time software systems with tight performance requirements are abundant. These systems frequently use many different algorithms and if these algorithms were to experience worst case behavior, the system may not be able to meet its performance requirements. Unfortunately, the inputs which would cause worst case behavior are often unknown, making it very difficult to defend against them. In this thesis, I present a method for finding the worst case inputs to different algorithms using simulated annealing, a combinatorial optimization method. I show that this method is successful in finding worst case inputs to several sorting algorithms, using several measures of an algorithm’s behavior.