Performance analysis and evaluation of postquantum secure blockchained federated learning

Abstract

As the field of quantum computing progresses, traditional cryptographic algorithms such as RSA and ECDSA are becoming increasingly vulnerable to quantum-based attacks, underscoring the need for robust post-quantum security in critical systems like Federated Learning (FL) and Blockchain. In light of this, we propose a novel hybrid approach for blockchain-based FL (BFL) that integrates a stateless signature scheme, such as Dilithium or Falcon, with a stateful hash-based scheme like XMSS. This combination leverages the complementary strengths of both schemes to provide enhanced security. To further optimize performance, we introduce a linear formula-based device role selection method that takes into account key factors such as computational power and stake accumulation. This selection process is reinforced by a verifiable random function (VRF), which strengthens the blockchain consensus mechanism. Our extensive experimental results demonstrate that this hybrid approach significantly enhances both the security and efficiency of BFL systems, establishing a robust framework for the integration of post-quantum cryptography as we transition into the quantum computing era.