ScholarWorks@숙명여자대학교

초록

Recentlry, the application of deep learning to assess source code similarity has been an area of interest. Although this approach is very promising in that it can improve the accuracy and efficiency of similarity detection tasks by leveraging the ability of deep learning models to automatically learn and extract features from various representations of source code, similarity assessment using deep learning has issues regarding the accuracy, fairness, and interpretability of the results. This paper considers the issues related to deep learning-based software similarity assessment techniques and discusses the improvements required for deep learning models to be used in software assessment from a technical perspective. This study aims to provide a practical method for deep learning-based software similarity assessment to be industrially efficient while minimizing legal disputes. Suggested improvements include introducing a hybrid approach, automating dataset augmentation and labeling, lightweighting and efficient learning of models, and introducing explainable AI. In addition, we present scenarios for utilizing various evaluation indicators to improve reliability.

키워드