In an initial attempt to systematize the research field of architectural threat analysis, this paper presents a comparative study of two threat analysis techniques. In particular, the controlled experiment presented here compares two variants of Microsoft’s STRIDE. The two variants differ in the way the analysis is performed. In one case, each component of the software system is considered in isolation and scrutinized for potential security threats. In the other case, the analysis has a wider scope and considers the security threats that might occur in a pair of interacting software components. The study compares the techniques with respect to their effectiveness in finding security threats (benefits) as well as the time that it takes to perform the analysis (cost). We also look into other human aspects which are important for industrial adoption, like, for instance, the perceived difficulty in learning and applying the techniques as well as the overall preference of our experimental participants.
