Cross-Language Source Code Clone Detection Using Deep Learning with InferCode
share
Software clones are beneficial to detect security gaps and software maintenance in one programming language or across multiple languages. The existing work on source clone detection performs well but in a single programming language. However, if a piece of code with the same functionality is written in different programming languages, detecting it is harder as different programming languages have a different lexical structure. Moreover, most existing work rely on manual feature engineering. In this paper, we propose a deep neural network model based on source code AST embeddings to detect cross-language clones in an end-to-end fashion of the source code without the need of the manual process to pinpoint similar features across different programming languages. To overcome data shortage and reduce overfitting, a Siamese architecture is employed. The design methodology of our model is twofold – (a) it accepts AST embeddings as input for two different programming languages, and (b) it uses a deep neural network to learn abstract features from these embeddings to improve the accuracy of cross-language clone detection. The early evaluation of the model observes an average precision, recall and F-measure score of 0.99, 0.59 and 0.80 respectively, which indicates that our model outperforms all available models in cross-language clone detection.
READ FULL TEXT