Improving Security for Facial Image Using Fragile Digital Watermarking

Abstract

Recently, face recognition technique is an active field (Zhao et al., 2000; Viola et al., 2001; Gu et al., 2002) in computer vision and anti-terror, safety defense, etc. Current face recognition pays more attention to these fields: 1) detection failure of some or all of features due to a variety of lighting conditions and head motions; 2) multiple and nonrigid object tracking; and 3) features occlusion when the head is in oblique angles and so on. But the security of facial image database has rarely been studied in face recognition system. Once hackers attack facial images database by unlawful approach, such as tamper, substitution and addition etc, there are great fatal influences to the availability of commercial face recognition systems. It is very important to improve the security of facial image database. Current security of facial image database depends on the conventional database protection strategy (Zhang et al., 2005; Su et al., 2005). To resist various attacks on facial image database, such as tamper, substitution and addition form unlawful approach, a fragile digital watermarking method is proposed to improve security of facial image database in this paper, which can distinguish any minute tampers on facial image, and detect the modification location. First, the low-frequency compressed facial image, which lowfrequency wavelet coefficients of 7 Most Significant Bits (MSBs) of original facial image are non-uniform scalar quantization, is converted into a binary sequence as the watermark to be embedded. The improved security watermark scrambled by chaotic systems is embedded into the LSB of the facial image data. When facial image is identified, the fragile digital watermarking method is able to detect the tampered location and discriminate the validity of facial image database, which comes from the difference facial image between the lowfrequency compressed image and reconstructed image by watermark. We also analyze the error estimation factor for identifying facial images embedded fragile digital watermarking. Experimental results show that the fragile digital watermarking technique has high sensitivity to tampers on the watermarked facial images, it not only can improve the security of the watermarked facial image database, but also will not impact feature extraction, detection rate and detection speed of face recognition. The organization of the paper is as follows: application of the fragile digital watermarking technique to facial image database is given in Section 2. Section 3 gives the theoretical analysis about the error estimation factor after embedded fragile digital watermarking. The