Comments on “Lightweight Privacy and Security Computing for Blockchained Federated Learning in IoT”

Federated learning has been used to collaboratively train a decentralized model without sharing confidential model records. However, many security risks are involved in this regard, and scholars have conducted numerous studies on related topics. Fan et al. proposed a scheme to achieve model unforgeability and confidentiality in blockchained federated learning (lightweight privacy blockchained federated-learning (LPBFL) scheme) in the Internet of Things. Our research demonstrates that their scheme is insecure in terms of signature design and that their theorem is invalid. We present an effective attack on their signature algorithm and create a new signature method and a formal security model to provide security guarantee against the mentioned attack. Extensive simulations demonstrate that our signature algorithm does not harm the highly efficient LPBFL scheme.