Optimized Deep Neural Model-Based Intrusion Detection and Mitigation System for Vehicular Ad-Hoc Network

Abstract

In this research work, a novel Intrusion Detection and Mitigation System (IDMS) is modeled based on the optimization assisted deep learning technique. The two major phases are: feature extraction, attack detection and mitigation. In the feature extraction phases, the features related to the traffic flow and the vehicle position gets extracted. Then, these extracted features are subjected to attack detection phase, where the weight optimized Deep Neural Network detects the presence/absence of attack in the network. For the tuning purpose, a new Improved Particle Swarm Optimization (IPSO) algorithm is introduced in this work. The presented method overcomes the traditional PSO's drawbacks, such as easily falling into local optima, and improves its performance. Once the attack behavior is identified, it is very important to mitigate the attacker from the network. For this, BAIT based mitigation process is used in this work. Finally, the performance of the proposed IDMS is evaluated over the extant techniques in terms of certain performance measures. The proposed work’s UR is 13% better than the current SVM, 9% better than PSO + DNN, 6.6% better than NB, 8.6% better than I-GHSOM, and 16.6% better than SLnO + DNN.

Keywords:

• BAIT framework
• intrusion detection
• IPSO
• mitigation
• optimized NN
• VANET