云计算辅助的高效决策树隐私保护查询协议

Efficient privacy-preserving decision trees evaluation protocol with cloud-assisted computing

针对基于线性函数的决策树隐私保护查询协议使用单比特同态加密比较算法,导致客户端的计算和通信开销较大问题,提出一种云计算辅助的高效决策树隐私保护查询协议。利用改进的双重ElGamal同态加密算法,将客户端与决策服务器之间执行的整数比较协议和决策节点选择协议转移至决策服务器与云计算辅助服务器之间,减少客户端的计算开销及其与服务器之间的通信开销,同时保护客户端的特征值信息和决策服务器的决策树模型信息的隐私。在乳腺癌数据集上的实验结果表明,该协议的客户端计算时间比基于线性函数的决策树隐私保护查询协议减少约41%,并且客户端与两个服务器之间的通信量减少约53.5%。

Linear function based privacy-preserving decision trees evaluation protocol applies an additive homomorphic bit-encryption scheme to realize secure integer comparison,which makes the computation and communication cost of the client is very large.A cloud-computing assisted privacy-preserving decision trees evaluation protocol is proposed.The integer comparison protocol and the decision node selection protocol between the client and the decision server are transferred to execute between a cloud-computing assisted server and the decision server by an improved double ElGamal homomorphic bit-encryption algorithm,which reduces the client's computation and communication overheads with the servers,and protects the eigenvalue information privacy of the client,as well as the decision tree model information of the decision server.Experimental results on data of breast cancer show that the client computing time of the proposed protocol is 41%less than that of the linear function based protocol.The communication between the client and the two servers is 53.5%less than that of linear function based protocol.