The operations research journal Interfaces is publishing an entire issue of homeland security-related articles this month, and highlighted one of the articles in it in a press release in mid-November. That article, entitled “How Effective is Security Screening of Airline Passengers?” attempts to answer a question that has long bedeviled TSA and other aviation screening agencies: what is more effective, risk-based screening or random screening?
The authors, Susan Martonosi at Harvey Mudd and Arnold Barnett at MIT, build a model that includes variables for primary screening effectiveness, secondary screening effectiveness, % of passengers elected for random screening, the effectiveness of the prescreening system, and the terrorists’ deterrence threshold. They test the model several times using different assumptions for these variables, and get results for a range of scenarios that suggest that neither risk-based secondary screening nor random secondary screening is inherently better than the other:
As this simple mathematical model suggests, neither side has made a persuasive case about the effectiveness of airport passenger-profiling systems. Supporters of such systems have focused mostly on the ability of the algorithm to identify terrorists (C), an ability they may well overestimate. They say little about screening effectiveness of both low-risk passengers and selectees (p1) and (p2), yet these effectiveness parameters are crucial to the overall success rate of the system. Skeptics may have given insufficient weight to deterrence (Ï„), because of which, the selection and screening system might prevent attacks even though it falls well short of perfect. Probing the system, as we have seen, could sometimes prevent a terrorist act rather than ensure its success.
They conclude the article by suggesting that improving the baseline screening for all passengers might be a better investment than improving prescreening or secondary screening. Overall, a good article, and a useful exercise in trying to analyze an issue too often guided by intuition or emotion. You can see the descriptions of some of the other articles in this issue of Interfaces at this link.