Automated Software Vulnerability Assessment with Concept Drift

Abstract

Software Engineering researchers are increasingly using Natural Language Processing (NLP) techniques to automate Software Vulnerabilities (SVs) assessment using the descriptions in public repositories. However, the existing NLP-based approaches suffer from concept drift. This problem is caused by a lack of proper treatment of new (out-of-vocabulary) terms for the evaluation of unseen SVs over time. To perform automated SVs assessment with concept drift using SVs' descriptions, we propose a systematic approach that combines both character and word features. The proposed approach is used to predict seven Vulnerability Characteristics (VCs). The optimal model of each VC is selected using our customized time-based crossvalidation method from a list of eight NLP representations and six well-known Machine Learning models. We have used the proposed approach to conduct large-scale experiments on more than 100,000 SVs in the National Vulnerability Database (NVD). The results show that our approach can effectively tackle the concept drift issue of the SVs' descriptions reported from 2000 to 2018 in NVD even without retraining the model. In addition, our approach performs competitively compared to the existing word-only method. We also investigate how to build compact concept-drift-aware models with much fewer features and give some recommendations on the choice of classifiers and NLP representations for SVs assessment.

An Investigation into Inconsistency of Software Vulnerability Severity across Data Sources

“…One such task that has been regularly investigated is severity prediction [5]. selected stage for prediction, e.g., for bug reports [9] or SV databases [18]. Through this RQ, we display the impacts on prediction performance that dataset selection can have, and hence motivate researchers to properly consider this issue.…”

“…Older users may be better at assessing SVs. data as input [9], [18], to predict the normalized severity categories described in Table IV. Following these practices, we preprocessed text descriptions through removal of stop words (using the NLTK and sklearn stopword list) and punctuation, conversion to lowercase, and stemming.…”

“…Following these practices, we preprocessed text descriptions through removal of stop words (using the NLTK and sklearn stopword list) and punctuation, conversion to lowercase, and stemming. The descriptions were then encoded using a bag-of-words model; we only extracted features for words that appeared in more than 0.1% of all descriptions [18]. We evaluated the same classifiers and tuned the same hyperparameters that were experimented with by prior works [18].…”

“…The descriptions were then encoded using a bag-of-words model; we only extracted features for words that appeared in more than 0.1% of all descriptions [18]. We evaluated the same classifiers and tuned the same hyperparameters that were experimented with by prior works [18]. Full details of the experimental setup are available in our reproduction package [10].…”

“…Full details of the experimental setup are available in our reproduction package [10]. Time-based validation methods have been shown to be important for this line of research [18]. Hence, we sorted our dataset by submission date of the report description, and then divided the dataset into a training, validation and test set through an 80:10:10 split.…”