Blockchain-based Decentralized Trust Aggregation for Federated Cyber-Attacks Classification in SDN-Enabled Maritime Transportation Systems

Abstract

Maritime transportation systems (MTS) based on software-defined networks (SDN) offer reliability, efficiency, and flexibility in network connection and configuration. However, these abilities make an SDN-based MTS with a single controller vulnerable to cyber-attacks. A centralized-based intrusion detection system (IDS) technique requires uploading the network traffic features as well as training data to the cloud server. An SDN-based MTS that is concerned with data privacy makes it challenging. This study proposes a blockchain-based federated cyber-attacks classification framework to maintain data in a decentralized manner, raise privacy concerns, and perform a secure decentralized aggregation mechanism. A proof-of-authority (PoA)-based blockchain network with an interplanetary file system (IPFS) as off-chain distributed storage is implemented to provide a secure and trusted federated IDS framework. Moreover, the proposed lightweight CNN-based IDS model used a residual connection and factorized convolution structure to fix the gradient vanishing issue and provide a low-complexity network structure. The model measurement shows that the proposed model performs better than state-of-the-art models in terms of accuracy, computing cost, and model design complexity.