基于自适应非均匀量化的安全人脸识别方法

随着人脸识别技术的广泛应用,相应的隐私及安全问题也引起了人们的主意。主成份分析是人脸识别技术最典型的算法之一。分析了主成份分析方法的安全性,针对其存在的安全漏洞,提出了一种基于自适应非均匀量化的安全人脸识别算法,将主成份分析方法得到的人脸模板特征进行量化,用量化特征的哈希值代替人脸模板特征存储在数据库中,并给出了该算法的安全性分析。实验结果表明该算法在不改变算法准确性的同时也提高了系统的安全性。

A secure face recognition algorithm based on adaptive non-uniform quantization

With the development of face recognition technology, some privacy and security problems arise. PCA was one of the popular approaches for face recognition. Aiming at the leakage for the security of PCA, a secure face recognition algorithm was proposed, based on adaptive non-funiform quantization which quantizes the template feature obtained by PCA and stores the hash value of the quantization feature instead of the template feature. Experimental results show this algorithm improved the security of the system while the accuracy was the same as PCA.