Mining Method of Code Vulnerability of Multi-Source Power IoT Terminal Based on Reinforcement Learning

With the rapid development of IoT, many heterogeneous power terminals are connected, substantially increasing the difficulty of network attack protection. How to ac-curately grasp the supporting techniques such as sample generation, vulnerability targeting, and vulnerability cor-relation in intelligent vulnerability mining to improve the efficiency of vulnerability mining and ensure grid security is a major challenge we are currently facing. This pa-per studies the multi-source power IoT terminal code vul-nerability mining method based on reinforcement learn-ing. Firstly, the static analysis method is used to scan and analyze the source code of the multi-source power IoT terminal, and the abstract syntax tree of the code is constructed. A bidirectional search path algorithm is adopted to expand the path range of the directed graph of the multi-source power Internet of things terminal. Sec-ondly, the vulnerability of multi-source power IoT ter-minal code is located by the concept of dynamic taint tracking. The taint tracking results are input into the deep neural network model as samples. Finally, the pro-tocol vulnerability mining model based on reinforcement learning is constructed to obtain the vulnerability mining results. The experimental results show that the method has high vulnerability mining accuracy and coverage, can record the type and location of vulnerabilities, and gen-erate vulnerability reports to improve the security of the smart grid. © (2023), All Rights Reserved.