It is very important to detect source code similarity in the field of both software engineering and computer science education. Source code stealing and copyright dispute in software engineering, as well as plagiarism detection of student assignment in programming course, call for automation of source code similarity detection. Reliable software to detect source code similarity also helps in the field of source code clustering and source code searching in wide range. Though such problem rises along with the invention of program language, there is not satisfying research result up till now. Source code has its particular structure, making it improper to use traditional pure text similarity detection method over it. This paper first introduce the history and current progress of this problem, analysis the difficulty, then presents a new system for source code similarity detect base on control flow graph analysis, along with the algorithm it uses and the implementation details. In this paper we also discuss the advantage and disadvantage of the method and ways of improve it.