A Novel Intrusion Detection Technique Based On Fog Computing Using Cholesky Factorization Based Online Sequential Extreme Learning Machines with Persistent Regularization

Abstract

Each end every aspect of life is penetrated by Internet of things (IoT) with high challenges of security. Cyber-attack scope in IoT increased exponentially with a growth in IoT connected devices. So, it makes a necessity to develop an efficient intrusion detection system (IDS) which should be fast, dynamic and scalable in IoT environment.

Based on fog computing, this paper proposes a novel intrusion detection technique which used Cholesky Factorization Based Online Sequential Extreme Learning Machines with Persistent Regularization (CF-OSELM-PRFF). From IoT traffic, attacks can be interrupted effectively by this approach. For extreme learning machine, equation is constructed recurrently by CF-OSELM-PRFF. In every cycle, Cholesky factorization is sued to solve the equation effectively.

Samples timeliness is dealt by CF-OSELM-PRFF using forgetting factor. Cost function in its regularization term works perfectly. With variable or fixed size of node, CF-OSELM-PRFF learns the data by chunk-by-chunk or one-by-one manner. Local fog nodes are distributed with centralized cloud intelligence in detecting attack, in order to detect attacks in a faster rate. Distributed intrusion detection mechanism is enabled by distributed architecture of fog computing. It includes interoperability, flexibility and scalability. With respect to accuracy of detection and response time, better results are produced by proposed system as shown by analysis.