Privacy-preserving decentralized parameter estimation in edge computing networks

Abstract:
Edge computing is an effective method for decentralized parameter estimation in wireless sensor networks where part of data fusion tasks are implemented at the edge side. In some privacy-sensitive scenarios, the sensors’ observations should not be leaked to adversaries. Since edge computing nodes do not have the same level of information privacy preserving mechanism as fusion center, decentralized estimation has higher risk of information leakage. In this paper, we consider the decentralized parameter estimation with best linear unbiased estimator, and design the privacy-preserving scheme based on the homomorphic public key encryption techniques. The edge computing nodes exchange ciphertext with wireless sensor nodes, and the part of data fusion tasks are implemented at the edge side directly with ciphertext. Fusion center recover the estimated value with the aggregated data from edge computing nodes. The privacy-preserving can be enhanced since ciphertext is not decrypted at the edge side. Simulation results show that the proposed approach can achieve high accuracy of parameter estimation while enhancing the privacy-preserving.