用前馈神经网络对软件理解中函数调用序列的混沌识别

A Detection of Chaos Function Transfer Series in Software Understand Based on Feedforward Neural Networks Approach

对有噪声小数据量时间序列的混沌识别,是目前国内外许多应用领域研究的热点与难点。利用BP神经网络的非线性函数逼近能力,对小数据有噪声的时间序列计算最大李亚谱诺夫指数,可判断该序列是否存在混沌现象。本文首创将这一算法经转换应用到软件逆向工程过程的分析中,结果表明,软件逆向工程过程分析中出现的函数(或类)调用序列有些存在、有些不存在混沌现象,这为理解软件系统构建高层结构和抽取重用信息而开发新方法与新技术找到了理论依据。

To detection of chaos in noise short series, it is research of hotspot and difficulty at present. Make use of ANN to detect chaos in short series, the method computes thd Lyapunov exponent estimation when a high level of noise is presented. This paper originally brings forward the algorithm and applies it to analysis in software reverse engineering process by correspond conversion, two examples prove that some function（or class） transfer sequences appearing in the software reveres engineering analysis have the chaos phenomena, some don＇t. It is very important for us to understand software of high constructions and take reuse information.