A Novel Privacy Protection Scheme for Iris Identification
Zhang Lei,Yu Lili,Wang Bin,Bian Xingchao
DOI: https://doi.org/10.1007/s11277-019-06688-1
IF: 2.017
2019-01-01
Wireless Personal Communications
Abstract:With the development of iris technology, iris identification has been widely used in various domains, as an efficient security authentication pattern. However, as more and more people have been involved in this authentication, they have begun to realize that this pattern may also lead to privacy leakage during the procedure of security authentication. For one thing, the adversary can reverse the iris information of the user with the unauthorized iris template and gets the privacy. For another, the adversary can also get the iris template from other poorly managed application and utilizes this template to access the user’s information in a special application to get the privacy of him. These two types of attack have seriously affected the promotion of iris identification. Thus, in order to cope with above attacks, and based on the conception of random projection and differential privacy protection, this paper proposes a novel privacy protection scheme short for RPDPP to protect the privacy of the user during the procedure of iris identification. In this scheme, a random projection is used in the phase of iris collection and template production, so that the iris template cannot be reversed into the iris information. Then before conserving the real template, several dummy templates that similar with the real iris template have been produced, and these dummies have disturbed the real template to comply with the restriction of ɛ -differential privacy. Thus, the real template of the user is difficult to be identified with other similar dummies, and the adversary cannot get the privacy of the user through multiple queries, even though he had obtained the iris template from other applications. At last, several experiments have been given with different parameters, and the protection ability has been compared with some similar algorithms, and then results of these experiments had further demonstrate the superiority of our proposed scheme.