Probability-Based Parameter Selection for Black-Box Fuzz Testing

Dynamic, randomized-input functional testing, or black-box fuzz testing, is an effective technique for finding security vulnerabilities in software applications. Parameters for an invocation of black-box fuzz testing generally include known-good input to use as a basis for randomization (i.e., a seed file) and a specification of how much of the seed file to randomize (i.e., the range).This report describes an algorithm that applies basic statistical theory to the parameter selection problem and automates selection of seed files and ranges. This algorithm was implemented in an open-source, file-interface testing tool and was used to find and mitigate vulnerabilities in several software applications. This report generalizes the parameter selection problem, explains the algorithm, and analyzes empirical data collected from the implementation. Results of using the algorithm show a marked improvement in the efficiency of discovering unique application errors over basic parameter selection techniques.