Orientation Range for Transposition According to the Correlation Analysis of 2DPalmHash Code

Abstract

Recently two-dimensional (2D) Palm Hash Code was proposed as a cancelable palm print coding scheme for secure palm print verification. However, there is vertical correlation between the vertically adjacent entries of 2DPalmHash Code. Armed with the statistical correlation analysis, an attacker can break into the palm print verification system easier than using brute-force attack. This paper analyzes the relation between the vertical correlation of 2DPalmHash Code and the corresponding vertical correlation of Gabor feature matrix. To suppress the vertical correlation of 2DPalmHash Code, both the real part and imaginary part of Gabor feature matrices, which are extracted along the orientations approaching vertical orientation, are transposed, respectively. The orientation rang suitable for transposition for vertical correlation suppression is determined by both correlation analysis and experimental results.