Recode-Decode-and-Compare: An Efficient Verification Scheme for Coded Edge Computing Against Collusion Attack

Recently, coded computing has great breakthroughs in edge computing, which can support and optimise many latency-sensitive and computation-intensive applications. In edge computing, a practical issue is that, comparing to computing devices configured in the cloud data center, edge devices may be unreliable and untrustworthy. Since coded computing has potential in speeding up the computation and protecting the confidentiality of computing data, coded edge computing (CEC) emerges. For a major security issue: verify the correctness of computation result in CEC, the existing efficient work performs verification by comparing multiple final results. The attackers (malicious edge devices) studied before are independent, while collusion attack is the realistic and important scenario in distributed computing and edge computing. To improve the security of CEC, in this paper, we study the verification issue of collusion attack and design the efficient verification scheme: Recode-Decode-and-Compare (RDC). By re-encoding the encoded data, collusive attackers cannot infer the linear relationship between different intermediate results. Therefore, the user can verify the correctness by comparing multiple final results. We also conduct solid theoretical analyses to show the successful verification probability and verification overheads. Finally, extensive simulation experiments demonstrate that the RDC verification scheme can efficiently verify the computation result and outperforms the state-of-the-art works.