What If We Were Wrong? Security Algorithms' Sensitivity to Noise in Terms of Defender Expected Utility

Overview

Objectives/Goals In security games, defenders allocate resources then the attacker attacks as to maximize utility. However in a given security game, if the matrix changes and a target is worth more or less than previously predicted, a recipe for disaster is created. Applied to the real world, smugglers, poachers, terrorists etc., could make away with doing much more harm. Methods/Materials This project tests two algorithms DOBSS and Match in terms of robustness to noise, or their sensitivity to change, by testing them in four different matrices with various amounts of noise in the input and measuring the defender utility. If two programs DOBSS and Match were tested on their sensitivity to noise in the matrix, it was believed that Match would yield a higher utility than DOBSS. Results Contrary to the hypothesis, on average between all matrices, DOBSS yields a higher expected utility of 1.33462 points,than match yielding a utility of 0.1992 points, an average 1.13542 points higher. DOBSS therefore within the given setting is more robust to noise than match. Conclusions/Discussion Alone, the results are not sufficient enough alone to indicate whether or not DOBSS is more or less robust to noise than Match, further research and more tests are needed to provide an answer that would hold true in almost all scenarios. However this project provides massive steps towards future projects, and sheds needed light on the issue.

Competition history

• CSEF 2013 Mathematics & Software · Entry J1414

Resources

• CSEF project summary ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Computational Sciences projects
• CSEF Computational Sciences projects
• Every CSEF 2013 project