An Uneven Distributed System for Dynamic Taint Analysis Framework

Dynamic taint analysis has been widely used in software testing, debugging, vulnerability detection and other fields. A popular idea is that we can combine dynamic taint analysis with symbolic execution techniques or fuzz techniques forming the testing framework to test automatically. When testing large applications which costs longer time, a distributed system can be very practical. However, the common distributed system is load balancing which distributes tasks without considering the various performance of each machine, resulting that some machines with poor configuration will burden too much load. In this paper, we present an uneven distributed system, which splits the dynamic taint analysis framework into some modules, and then distributes the modules to different machines classified by their performance. The design and distribution method are all based on the feature of each module. In the studies, we applied the system to test 5 applications compared with the load balancing distributed system, and the results shows it can indeed distribute tasks uneven according to different performance.