Linear Model Against Malicious Adversaries with Local Differential Privacy
This addresses the problem of enabling efficient and secure data analysis for scientific collaborations with sensitive data, though it appears incremental by building on existing privacy-preserving techniques.
The paper tackles secure collaborative learning against malicious adversaries by applying matrix encryption to achieve local differential privacy and resist attacks like chosen plaintext and collusion, while maintaining computational efficiency. Empirical results on real-world datasets show the proposed schemes are more efficient than existing techniques for both malicious and semi-honest models.
Scientific collaborations benefit from collaborative learning of distributed sources, but remain difficult to achieve when data are sensitive. In recent years, privacy preserving techniques have been widely studied to analyze distributed data across different agencies while protecting sensitive information. Most existing privacy preserving techniques are designed to resist semi-honest adversaries and require intense computation to perform data analysis. Secure collaborative learning is significantly difficult with the presence of malicious adversaries who may deviates from the secure protocol. Another challenge is to maintain high computation efficiency with privacy protection. In this paper, matrix encryption is applied to encrypt data such that the secure schemes are against malicious adversaries, including chosen plaintext attack, known plaintext attack, and collusion attack. The encryption scheme also achieves local differential privacy. Moreover, cross validation is studied to prevent overfitting without additional communication cost. Empirical experiments on real-world datasets demonstrate that the proposed schemes are computationally efficient compared to existing techniques against malicious adversary and semi-honest model.