Feedback Driven Grammar-Based Fuzzing

Abstract

In this paper, we present a method for grammar-based fuzzing, which improves its penetration power. It is based on input data generation using a fuzzer feedback. Several other methods are prone to create an initial set of acceptable test cases before the actual fuzzing process, and hence are unable to use the runtime information to increase the generated input’s quality. The proposed method uses the coverage information gathered for each input sample and guides grammar-based input generation. This method uses more than 120 BNF (Backus-Naur Form) grammar rules described in ANTLR (Another Tool for Language Recognition) platform. Experimental results show that our method - feedback driven random test generation, has higher code coverage capabilities compared with the existing methods.