Multi-objective optimization of federated intrusion detection system for detecting black hole and gray hole attacks in MANET

Mobile Ad Hoc Networks (MANETs) are networks that do not have a permanent infrastructure, and are vulnerable to attacks like Black Hole Attacks (BHA) and Gray Hole Attacks (GHA) in which packets are dropped by rogue nodes. While trustworthiness of nodes is important, the traditional approaches to preventing this do not always succeed due to poor detection methods. To overcome these limitations, a novel framework is proposed to combine multi-objective optimization (MOO), Long Short-Term Memory (LSTM) based dynamic trust prediction, and federated learning (FL) for decentralized trust model updates with privacy preservation on Ad hoc on demand Routing protocol (FLT-AODV). In the approach, LSTM is used to learn temporal behavioral patterns to accurately evaluate trust and NSGA-II is used to optimize multiple routing goals: trust, energy efficiency and latency. The decision of adaptive routing is based on real-time trust evaluation and network conditions. The simulation results show that critical performance metrics have improved significantly with the highest Packet Delivery Ratio (PDR) of 99.9%, the highest throughput of 99.26 Kbps and the minimum E2E delay of 0.035 s. The methodology ensures optimal security by choosing the suitable nodes to send data, which helps in reducing BHA and GHA threats.