Fast Comparison of Software Birthmarks for Detecting the Theft with the Search Engine

The software birthmarks were proposed for detecting the software theft, which is suspected a copy. The birthmark system extracts native characteristics of a program. Then, the system computes the similarity between two birthmarks. This paper proposes the reduction method of the comparison time for the birthmark system. The conventional birthmark systems did not care the comparison time. Therefore, the enormous amount of programs could not compare practically for requiring quite a huge comparison time. To solve the problem, we tackled to reduce the comparison time with the search engine. So that we introduce the pre-comparison phase into conventional birthmark procedures to narrow the defendant programs. We implemented the proposed method with Apache Solr search engine and evaluated the method. The result was shown that the method for UC birthmark dramatically reduces the comparison time 81.35%, and narrows the target programs 74.81%. On the other hand, the results were not good in the default threshold ε = 0.75 (high false negative rates C_FN = 67.01%, and low accuracies M_a = 34.18%). However, the results were improved C_FN = 21.17% and M_a = 78.56% in average by lowering ε = 0.25.