Estimation of Probability of Adversary Mission Success

Historically, engineered security systems have been developed and optimized using variants on the formula for Probability of System Effectiveness, P(E) = P(I)*P(N) where P(I) is the Probability of Interruption (dependent upon a Response Force Time) and P(N), which is the Probability of Neutralization, given Interruption. This equation has allowed engineers to successfully determine how much delay was required after a perimeter with sensors, assessment, entry control, and screening points based on the requirement to exceed the response time. Security forces could then develop plans within these time and space constraints that would neutralize the adversary. This formula has two limitations: 1) P(I)*P(N) is a conservatively low bound for P(E) and 2) the formula models a single team of attackers. In today’s environment there is a need to extend and correct this equation, both to reflect effectiveness more accurately in P(E) and to properly model multiple team attacks. This paper discusses how to estimate the Probability of Adversary Mission Success, P(AMS), from the adversary’s perspective, which we believe can address some of these limitations. While it is common to use 1-P(E) as an estimate of P(AMS), we suggest that, in many cases, this is indeed an overestimate, that P(AMS) is much lower than 1-P(E). If true, this suggests that our systems will appear more robust in certain ways to an adversary planner than we give them credit for. Terms in P(AMS) but not in 1-P(E) include attack planners’ estimate of the probability that the multiple components of the attack can be successfully performed simultaneously and be synchronized properly, that the attack will not be compromised before it is carried out, and that attack essential data are correct. We will discuss how to estimate some of these terms using subject matter experts or performance testing.