Research On Program Properties Based Java Static Software Birthmarking

As the invariable features of a program, software birthmark can be used to detect software theft and protect the copyright of software. The software birthmarking for Java program is discussed in the thesis mainly, based on the brief introduction about the status of software birthmarking, some descriptions of novel software birthmarking are proposed. Research works in the thesis are the following three aspects:1) To improve the resilience of static k-gram software birthmark, the rate of change of k-gram frequencies between original program and transformed programs are computed based on counting frequencies of all k-grams of programs. Weights of all k-grams in original program are adjusted, thus a software birthmark based on weighted k-gram is proposed. Experimental results show that resilience of the proposed birthmark for Java applications is better than static k-gram birthmark and frequencies of k-gram based software birthmark. The performance of discrimination of same programs between different versions and same categories is keeping stability. Resilience of the proposed birthmark against code obfuscation and optimization is improved obviously, while credibility is little degeneration.2) A series of software birthmark based k-gram idea have low resilience performance to resist semantics-preserving transformations such as code rearrangement, redundant code insertion and so on. By introducing dataflow analysis techniques, dependencies of instructions that define or use variables in program are established. Then a novel method of software birthmarking based on use-define chains is proposed. The birthmark reflects software semantics to some extent and has high resilience performance to resist transformations that do not affect dataflow information of program. Similarity of methods enhances the performance of birthmark for module localization and detection.3) Software birthmark based single program property has limitation for describing program precisely, and detection threshold of similarity is uncertain as the choosing regular of artificial experience. Integrated with mutli-properties of program and machine learning method, a software birthmark based mutli-properties of program is proposed, then machine learning method of SVM is introduced for classifying. The new research idea extracts and analyses properties from different level of program, and program is comprehensive described. Machine learning method of SVM is used to overcome the artificial experience for choosing similarity comparison model and detection threshold.Finally, sum up our work and put forward the future research.