An effective fuzz input generation method for protocol testing

Abstract

Random fuzzer is a powerful tool to find vulnerabilities of protocol implementations. One important approach to generate test inputs uses regular expression, and the input may include all possible combinations of fundamental characters. Thus the number of test cases may be tremendous, and furthermore, equivalent strings may usually be generated. In this paper, an effective fuzz input generation method integrating random signal process and regular expression is proposed. Random signal processing technology helps to reduce equivalent inputs, and the regular expression method helps to cover as much special cases in input domain. By enlarging the distance between input cases, the method can reduce equivalent inputs dramatically, while at the same time cover as much special inputs. Protocol specific information element is also considered in the method.