Biblio

The rapid rise of cyber-crime activities and the growing number of devices threatened by them place software security issues in the spotlight. As around 90% of all attacks exploit known types of security issues, finding vulnerable components and applying existing mitigation techniques is a viable practical approach for fighting against cyber-crime. In this paper, we investigate how the state-of-the-art machine learning techniques, including a popular deep learning algorithm, perform in predicting functions with possible security vulnerabilities in JavaScript programs. We applied 8 machine learning algorithms to build prediction models using a new dataset constructed for this research from the vulnerability information in public databases of the Node Security Project and the Snyk platform, and code fixing patches from GitHub. We used static source code metrics as predictors and an extensive grid-search algorithm to find the best performing models. We also examined the effect of various re-sampling strategies to handle the imbalanced nature of the dataset. The best performing algorithm was KNN, which created a model for the prediction of vulnerable functions with an F-measure of 0.76 (0.91 precision and 0.66 recall). Moreover, deep learning, tree and forest based classifiers, and SVM were competitive with F-measures over 0.70. Although the F-measures did not vary significantly with the re-sampling strategies, the distribution of precision and recall did change. No re-sampling seemed to produce models preferring high precision, while re-sampling strategies balanced the IR measures.

Background: Bug datasets have been created and used by many researchers to build bug prediction models. Aims: In this work we collected existing public bug datasets and unified their contents. Method: We considered 5 public datasets which adhered to all of our criteria. We also downloaded the corresponding source code for each system in the datasets and performed their source code analysis to obtain a common set of source code metrics. This way we produced a unified bug dataset at class and file level that is suitable for further research (e.g. to be used in the building of new bug prediction models). Furthermore, we compared the metric definitions and values of the different bug datasets. Results: We found that (i) the same metric abbreviation can have different definitions or metrics calculated in the same way can have different names, (ii) in some cases different tools give different values even if the metric definitions coincide because (iii) one tool works on source code while the other calculates metrics on bytecode, or (iv) in several cases the downloaded source code contained more files which influenced the afferent metric values significantly. Conclusions: Apart from all these imprecisions, we think that having a common metric set can help in building better bug prediction models and deducing more general conclusions. We made the unified dataset publicly available for everyone. By using a public dataset as an input for different bug prediction related investigations, researchers can make their studies reproducible, thus able to be validated and verified.