Federated Deep Learning-Based Intrusion Detection System for Multi-Attack Detection in MANETs

Abstract

Mobile Ad-Hoc Networks (MANETs) have become an essential component of modern wireless communication systems, especially in emergency response, tactical military environments, and mobile IoT applications. Their architecture, which operates without fixed infrastructure, makes them inherently flexible yet highly vulnerable to security threats. Malicious nodes can launch routing-based attacks such as blackhole, grayhole, wormhole, and Sybil, severely degrading network performance. Traditional Intrusion Detection Systems (IDS) are predominantly centralized, requiring global data aggregation and shared computational resources. This approach increases privacy risks, elevates latency, and fails to scale effectively in dynamic MANET topologies. This paper introduces a novel Federated Deep Learning-based Intrusion Detection System (FL-IDS) designed to detect multiple routing attacks in real-time within a decentralized MANET environment. Unlike centralized IDS models, the FL-IDS employs federated learning, in which each node trains a local deep neural network using its data and transmits only model parameters to a central aggregator for federated updates. The neural architecture integrates a convolutional neural network (CNN) with an autoencoder-based feature reduction module to enhance detection accuracy while minimizing communication overhead. Simulations are performed using NS-3 under various network and attack conditions. Experimental results indicate that the proposed FL-IDS achieves a detection rate of 97.9%, an accuracy of 98.7%, and a false positive rate of just 1.4%, outperforming conventional centralized IDS architectures. The proposed system demonstrates excellent scalability, low communication overhead, and high adaptability—making it a promising solution for secure MANET deployments in resource-constrained environments.

Introduction

Mobile Ad-Hoc Networks (MANETs) are decentralized, self-organizing wireless networks used in environments where fixed infrastructure is unavailable. Their dynamic and resource-constrained nature makes them highly vulnerable to routing attacks such as blackhole, grayhole, wormhole, and Sybil attacks. Traditional Intrusion Detection Systems (IDS), especially those relying on centralized data collection, struggle with MANET constraints including privacy risks, high communication overhead, latency, and poor adaptability to new attack patterns.

To address these challenges, recent research has explored machine learning and deep learning, but most approaches still depend on centralized training or focus on limited attack types. Federated learning (FL) offers a promising alternative by enabling distributed model training without sharing raw data. Although FL has been applied to IoT, VANETs, and limited MANET scenarios, prior work rarely supports multi-attack detection or optimized architectures.

The proposed study introduces a comprehensive Federated Learning-based Intrusion Detection System (FL-IDS) featuring a hybrid deep learning architecture combining an autoencoder for feature reduction with CNN layers for spatial–temporal pattern extraction. Each node locally monitors routing behavior, trains the model on its own data, and sends only model updates to a federated aggregator using FedAvg—preserving privacy and reducing communication load. The system is evaluated under multiple routing attacks, including combinations of attack types.

Simulation using NS-3 across various MANET sizes, mobility levels, and attacker densities shows that the FL-IDS significantly outperforms a centralized IDS. It achieves higher accuracy (98.7%), higher detection rate (97.9%), much lower false positive rate (1.4%), reduced detection latency (2.1 s), and moderate communication overhead. The architecture scales well, demonstrates robustness under multi-attack scenarios, and maintains efficiency on resource-limited nodes.

References

[1] Hussain, S. F. M., & Fathima, S. M. H. S. “Federated Learning-Assisted Coati Deep Learning-Based Model for Intrusion Detection in MANET.” International Journal of Computational Intelligence Systems, 2024, 17:285. DOI: 10.1007/s44196-024-00590-w. [2] Ennaji, M., & others. “Adversarially robust federated deep learning models for intrusion detection in IoT.” Indonesian Journal of Electrical Engineering and Computer Science, 2024, 37(2), pp. 937–947. DOI: 10.11591/ijeecs.v37.i2.pp937-947. [3] Tesfay, D., & others. “An Intrusion Prevention System embedded AODV to Detect and Prevent Sybil Attack in MANET.” ACM, 2021. DOI: 10.1145/3484824.3484915. [4] Nirmala Bai, K. S., & Subramanyam, M. V. “Integrated intrusion detection design with discretion of leading agent using machine learning for efficient MANET system.” Scientific Reports, 2025, 15:30849. DOI: 10.1038/s41598-025-89221-8. [5] Laqtib, S., “A deep learning methods for intrusion detection systems-based machine learning in MANET,” ACM, 2019. DOI: 10.1145/3368756.3369021. [6] Venkatesubramanian, S. , “Detection of Black Hole Attack Using Honeypot Agent-Based Scheme with Deep Learning Technique on MANET,” ISI Journals (IIETA), 2021, 26(6), pp. 549–557. DOI: 10.18280/isi.260605. [7] Venkateswaran, N., & Prabaharan, S. “An Efficient Neuro Deep Learning Intrusion Detection System for Mobile Ad-hoc Networks.” EAI, 2022. DOI: 10.4108/eai.4-4-2022.173781. [8] Sardar, T. H., “Enhancing Security in MANETs with Deep Learning-Based Graph Neural Network for Intrusion Detection,” ScienceDirect, 2025. DOI: 10.1016/j.procs.2025.1010579. [9] Shafi, S., “Machine Learning and Trust Based AODV Routing Protocol to Mitigate Flooding and Blackhole Attacks in MANET,” ScienceDirect, 2023. DOI: 10.1016/j.procs.2023.101431. [10] Khan, K., “A Survey on Intrusion Detection and Prevention in Wireless and Mobile Ad Hoc Networks,” Elsevier, 2020. DOI: 10.1016/j.comcom.2019.11.013. *(derived from survey topic) * [11] Amouri, A., “A Machine Learning Based Intrusion Detection System for Mobile Ad-Hoc Networks,” Sensors, 2020, 20(2). DOI: 10.3390/s20020461. [12] Anantvalee, T., & Wu, J. “A Survey on Intrusion Detection in Mobile Ad Hoc Networks,” in Advances in Wireless Ad-hoc Networks, Springer, 2009. DOI: 10.1007/978-0-387-33112-6_7. [13] Nallusamy, R., Jayarajan, K., & Duraiswamy, K. “Intrusion Detection in Mobile Ad Hoc Networks Using GA Based Feature Selection.” arXiv, 2009. DOI: 10.48550/arXiv.0912.2843. [14] Mitrokotsa, A., Tsagkaris, M., & Douligeris, C. “Intrusion Detection in Mobile Ad Hoc Networks Using Classification Algorithms.” arXiv, 2008. DOI: 10.48550/arXiv.0807.2049. [15] Khan, M., “A Comprehensive Survey on Machine Learning–Based Intrusion Detection Systems in IoT, MANET and Wireless Networks.” Wiley, 2023. DOI: 10.1155/2023/8981988.

Copyright

Copyright © 2025 Arun Joseph A, Ranjani P, Suresh Kumar V. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.