A FEDERATED LEARNING-BASED ATTACK DETECTION METHOD FOR BLOCKCHAIN-BASED POWER TRADING SYSTEMS

With the increase of the network scale, blockchain-based power trading systems are facing increasingly serious network attacks. Limited by departmental privacy policies, current research still uses data samples from a single department for attack detection, making it difficult to accurately perceive complex network attacks. In response to this issue, a federated learning-based attack detection method for blockchain-based power trading systems is proposed. The method can perform joint model training under the premise of ensuring the protection of multi-departmental data and identity privacy, thereby providing systems with more comprehensive detection capabilities. Experimental results show that the proposed method not only improves the accuracy of attack detection but also guarantees the privacy of various departments.