Privacy-Preserving Swarm Learning Based on Lightweight Homomorphic Encryption and Blockchain Technology

A decentralized solution called swarm learning integrates blockchain into machine learning with distributed data sources and computing power. The security of blockchain is inherited and the security and privacy of data is enhanced surpassing federated learning. Previous studies have explored the application of homomorphic encryption to privacy protection in distributed machine learning. However, existing homomorphic encryption algorithms require a large amount of computation, making them unsuitable for swarm learning. To solve this problem, we improved a privacy-preserving swarm learning framework based on an improved lightweight Paillier homomorphic encryption algorithm. We evaluated the performance of the improved algorithm under the federated learning environment and the security and the swarm environment. Experiment results showed that under the MNIST data set, the encryption time of the classical homomorphic encryption algorithm was 2.9 times longer than that of the improved algorithm. The improved algorithm improved the encryption efficiency of the homomorphic encryption algorithm and protected the privacy of swarm learning data.