Fed-ANIDS: Federated learning for anomaly-based network intrusion detection systems

Meryem Janati Idrissi, Hamza Alami, Abdelkader El Mahdaouy, Abdellah El Mekki, Soufiane Oualil, Zakaria Yartaoui, Ismail Berrada

Abstract: Highlights•Fed-ANIDS is an efficient method for distributed network intrusion detection.•Fed-ANIDS leverages the power of federated learning approach and autoencoders techniques.•Autoencoders perform better than GANs in distributed network intrusion detection.•FedProx outperforms FedAvg in a distributed network intrusion detection environment.•Extensive evaluations of Fed-ANIDS with USTC and CIC-IDS17/18 show promising results.