Answers are grounded in this paper and cite their sources.
Working for a strategy to ensure web application security has become more complex as it is not only to protect against the more challenging cross-site scripting (XSS) attacks but also to assist the open expressiveness of web applications to provide users with interactive services. The feature extractions used in supervised machine learning to detect XSS as an approach to the strategy are now in question of their effective classification. Their lack of preserving structural information may not uphold the property of structured data in input payloads. We apply the concept of syntactic n-grams to a payload text representation. The study of different feature extractions on the representation is to see if syntactic information is maintained. Our purpose is to determine the more effective approach to correctly classify benign and malicious payloads on a real-world dataset. The use of sn-grams that produces the most favourable results of accuracy and precision would only indicate that the extraction is reasonably able to minimize false reports during classification.
No takes yet. Share an insight, caveat, or question.
Talib et al. (2022) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: