ICPFuzzer: proprietary communication protocol fuzzing by using machine learning and feedback strategies

Abstract The fuzzing test is able to discover various vulnerabilities and has more chances hit the zero-day targets. And ICS(Industrial control system) currently facing huge security threats requires standards, like ISO 62443, ensure quality of device. However, some industrial proprietary communication protocols can be customized have complicated structures, system cannot quickly generate data that adapt protocols. It also struggles define mutation field without having prior knowledge Therefore, we propose a named ICPFuzzer uses LSTM(Long short-term memory) learn features protocol generates mutated automatically. We use responses testing adjust weight strategies further device under (DUT) find cause unusual connection status. verified effectiveness approach by comparing with open-source commercial fuzzers. Furthermore, in real case, experimented DLMS/COSEM for smart meter found response. In summary, black-box automatically execute process reveal interrupt crash communication. Not only improves ICS but safety.