恶意模型下隐私保护点积的研究

Research on Privacy Preserving Inner Product for Malicious Model

由于现存的隐私保护方法大多是在半诚实模型下或针对某一算法,且基于安全多方计算的算法效率较低,因此使用随机正交变换扰乱技术实现恶意模型下垂直分布数据隐私保护点积的计算,该方法能抵制恶意方的共谋,有较好的伸缩性,实现了数据的有效保护。理论证明和实验分析表明该方法的安全性和经过扰乱后数据的有效性。

The existing privacy preserving methods mostly based on semi-honest models and concentrate on a single algorithm,and secure multi-party computation are inefficient,so this paper proposes a privacy-preserving method based on random distribution for vertically partitioned data to compute inner product.The technique is suit for malicious models by choosing random matrix which according with some condition,and protects the inner product between different attributes effectively.Theoretic argument and example analysis demonstrate that our scheme is secure and maintain the validity of data.