【摘要】 在实际数据收集中,不同敏感值的敏感度有很大差异,隐私保护需求也不相同.然而,现有的基于随机响应的本地化隐私保护模型针对所有敏感值都执行同样程度的隐私保护,从而可能造成某些低敏感度的敏感值过度保护,而某些高敏感度的敏感值却保护不足.基于此,本文在常规随机响应(Conventional Randomized Response,CRR)模型的基础上,考虑个性化的隐私需求,引入敏感值权重,并将其引入到随机响应的决策中,提出一种面向多敏感值的个性化随机响应(Personalized Randomized Response,PRR)机制,该机制能够确保不同的敏感值群体均能达到各自期望的隐私保护程度,实现个性化的隐私保护.理论分析和仿真实验表明,在机制的主观隐私泄露程度一定时,相比于CRR模型,本文所提的PRR机制统计估计误差更小,即获得的统计数据的质量更高,同时又保证了个性化的隐私保护.更多还原

【Abstract】 In actual dada collection,the sensitivity of different sensitive information is different so that the concrete privacy need is different,too.However,the existing local privacy preservation model based on randomized response(RR),which is called conventional randomized response(CRR) for convenience,focuses on a universal approach that exerts the same amount of preservation for all sensitivity values,without catering for their concrete privacy needs.As a result,it may be offering insufficient protection to a subset of people with relatively higher privacy needs,while applying excessive privacy control to another subset with relatively lower privacy needs.Based on this,a new framework which is called personalized randomized response(PRR) is proposed based on the concept of CRR for multiple sensitive values-oriented personalized privacy preservation.The PRR technique considers personalized privacy needs,introduces sensitive value weights for different sensitive values,and then introduces the weights into the decision of RR for satisfying all sensitivity values’ privacy needs,and thus,attains personalized privacy preservation.Both theoretical derivation and simulation experiment reveal that the estimate error of statistics of PRR mechanism is smaller than that of the CRR mechanism for a certain subjective degree of privacy leakage,that is,the quality of statistics obtained by PRR mechanism is higher than that of the CRR model while guaranteeing personalized privacy protection for a given subjective degree privacy preservation.更多还原