Method for intelligently detecting C program memory leakage based on machine learning

Abstract

The invention discloses a method for intelligently detecting C program memory leakage based on machine learning. The method has three stages; in a model construction stage, two data sets are constructed according to original memory leakage; memory leakage characteristics are extracted respectively from the two data sets; the memory leakage characteristics are input into a machine learning classifier, so that training and cross validation are carried out; the type and the parameter of the classifier are modified; the memory leakage is detected by taking the classifier having highest classification accuracy rate as the classifier model; in a program analysis and characteristic obtaining stage, a source program is pre-analyzed; all memory allocation points o are obtained; then, pointer analysis is carried out; a VFG (Value Flow Graph) at the beginning of o is constructed; the memory leakage characteristics corresponding to each path in the VFG are extracted; in a defect detection and report stage, the memory leakage characteristics in the program analysis and characteristic obtaining stage are input into the classifier model in the model construction stage and detected; whether memoryleakage exists in a path at the beginning of o or not is judged; and thus, a memory leakage report is obtained.

Description

technical field [0001] The invention relates to a memory leak detection method, which uses a program static analysis method and a machine learning algorithm to realize memory leak detection, and belongs to the field of software engineering. Background technique [0002] In computer science, program static analysis refers to the analysis of computer software without actually executing the program. In most cases the analysis is performed on some version of the source code, in other cases it is some form of object code. Common machine learning algorithms include decision trees, random forests, logistic regression, SVM, naive Bayes, etc. The advantage of SVM is that it simplifies the usual classification and regression problems. The complexity of its calculation depends on the number of support vectors, not the dimension of the sample space. Therefore, SVM is not only simple in algorithm but also has good robustness. [0003] At present, the detection of memory leaks mainly ad...

AI Technical Summary

Problems solved by technology

[0004] The current memory leak detection methods usually use static analysis and dynamic detection. Static analysis has too many false positives, and cannot solve memory leaks caused by some complex structures in C language, and the overhead of dynamic detection is high.

Images