In response to the problem of balancing privacy protection and data utility in mobile group perception data auction scenarios, this study proposes Lightweight adaptive differential privacy protection mechanism based on sparse shared neighbors for density peak clustering LDCR(SNDPC-LDCR). Firstly, the GRU network is used to predict time-series data, calculate data change rate, and combine PID error with remaining privacy budget to achieve adaptive sampling; Secondly, the recycling factor is introduced to optimize the allocation of sliding window privacy budget, and SNDPC clustering is used to complete data grouping. Noise is applied according to the minimum budget within the group and smoothed by Kalman filtering; Finally, non-sampled data will be published using the previous results, balancing privacy strength and computational efficiency. The experimental results show that the proposed algorithm has lower MAE than the compared methods under different sliding window lengths and overall privacy budgets, with smaller data distortion and better usability; The allocation of privacy budget is more reasonable, with stronger robustness to changes in window and budget parameters, and significant lightweight characteristics. Research has shown that the SNDPC-LDCR mechanism can achieve a balance between privacy protection and data utility in mobile group aware data auctions, providing efficient technical support for secure data transactions and open sharing.
