Research Article
Deep Reinforcement Learning Based Mimicry Defense System for IoT Message Transmission
@INPROCEEDINGS{10.1007/978-3-030-97124-3_31, author={Zhihao Wang and Dingde Jiang and Jianguang Chen and Wei Yang}, title={Deep Reinforcement Learning Based Mimicry Defense System for IoT Message Transmission}, proceedings={Simulation Tools and Techniques. 13th EAI International Conference, SIMUtools 2021, Virtual Event, November 5-6, 2021, Proceedings}, proceedings_a={SIMUTOOLS}, year={2022}, month={3}, keywords={Mimicry defense Deep reinforcement learning IoT}, doi={10.1007/978-3-030-97124-3_31} }- Zhihao Wang
Dingde Jiang
Jianguang Chen
Wei Yang
Year: 2022
Deep Reinforcement Learning Based Mimicry Defense System for IoT Message Transmission
SIMUTOOLS
Springer
DOI: 10.1007/978-3-030-97124-3_31
Abstract
With the development of 5G and Internet of Everything, IoT has become an essential network infrastructure. The connection between massive devices brings huge convenience and effectiveness, also introducing more security threats and vulnerabilities that compromise the security, privacy and trust problem of the IoT data, devices and users or service providers. Traditional security approaches are mostly based on the analysis of attack characteristics, seeking vulnerabilities, or patching systems. Independent from prior knowledge or specific defense method, the mimic defense can realize a built-in security system through heterogeneity, redundancy, and dynamic. In this paper, to address the security problem of the IoT communication protocol MQTT, a DRL-based mimicry defense system for IoT message transmission is proposed. We conduct mimic transformation on the MQTT broker, with functionally equivalent but structural dissimilar variants. To refine the determining accuracy of basic mimic ruling mechanism, namely majority voting, an intelligent ruling mechanism based on deep Q network is proposed. Finally, the simulation results demonstrate the security and effectiveness of the proposed scheme.
