Abstract:

A texture-fingerprint-based approach is proposed to extract or detect the feature from malware content. The texture fingerprint of a malware is the set of texture fingerprints for each uncompressed gray-scale image block. The ma-licious code is mapped to uncompressed gray-scale image by integrating image analysis techniques and variants of mali-cious code detection technology. The uncompressed gray-scale image is partitioned into blocks by the texture segmen-tation algorithm. The texture fingerprints for each uncompressed gray-scale image block is extracted by gray-scale co-occurrence matrix algorithm. Afterwards, the index structure for fingerprint texture is built on the statistical analy-sis of general texture fingerprints of malicious code samples. In the detection phase, according to the generation policy for malicious code texture fingerprint, the prototype system for texture fingerprint extraction and detection is con-structed by employing the integrated weight method to multi-segmented texture fingerprint similarity matching to de-tect variants and unknown malicious codes. Experimental results show that the malware variants detection system based on the proposed approach has good performance not only in speed and accuracy but also in identifying malware variants.

Key words: network security, malware variants detection, texture fingerprint, spatial similarity retrieval