PEFL: A Privacy-Enhanced Federated Learning Framework for Mobile Edge CrowdSensing in the Presence of Collusion and Backdoor Att

This dataset is associated with the manuscript titled \u201cPEFL: A Privacy-Enhanced Federated Learning Framework for Mobile Edge CrowdSensing in the Presence of Collusion and Backdoor Attacks.\u201d It is designed to evaluate the effectiveness of the proposed PEFL framework under various attack scenarios in mobile edge crowdsensing environments. The dataset includes simulated local model updates generated from multiple participants, with labels indicating benign, backdoor-injected, and colluding behaviors. It also incorporates relevant metadata such as differential privacy parameters, clustering information, and encryption configurations used during the experiments. This dataset enables comprehensive testing of privacy-enhancing techniques, including Backdoor Resistant Privacy-Enhanced Aggregation (BRPEA) and Collusion Resistant Privacy-Preserving Aggregation (CRPEA), as well as comparisons with baseline methods. It supports reproducibility of the theoretical and experimental results reported in the manuscript.