Loose Game Theory Based Anomaly Detection Scheme for SDN-Based mMTC Services

Abstract

In this paper, we exploit the game theory to integrate the strengths of statistics-based detection and the machine learning-based detection for wireless software-defined-networking (SDN) based massive machine-type-communications (mMTC) services. Different from existing game theory approaches that require perfect rationality of the players and the synchronized information shared between the players, the proposed scheme is designed to bear with fewer assumptions. A novel feedback mechanism is proposed to fed machine learning-based detection results into strategy selection for advanced strategy decision making when the maliciousness estimation is not accurate. Besides, a multi-hop architecture is proposed to enable distributed detection instead of common cluster style for enhanced scalability. Various scenarios are set to testify the performance of the proposed scheme. Simulation results show that the proposed scheme relives more pressure of controller, consumes less energy without showing evident influence on the regular packets delivery, under the condition of asymmetric knowledge and multi attack-pattern.

Keywords

• Anomaly detection
• DDoS
• energy-efficient
• mMTC
• SDN