A modular, locally deployable and model agnostic framework that reviews PHP files with open source large language models instead of signatures or regular expressions. Built to stay private, since nothing needs to leave the server.
"}},{"node":{"frontmatter":{"order":2,"title":"Quality-Gated Hybrid Code Mutation for Enhanced PHP Malware Detection","authors":"Neupane, B., & Lim, K.","venue":"International Conference on Convergent and Smart Systems (ICCSS 2026)","place":"Nairobi, Kenya","published":"July 2026","pages":null,"doi":null,"url":null,"award":null,"awardUrl":null},"html":"A quality gated pipeline that pairs LLM code mutation with rule based mutation to build training data, then filters duplicates and near duplicates before training so the extra data does not just add noise.
"}},{"node":{"frontmatter":{"order":3,"title":"Machine Learning-Based Detection of PHP Web Shells: A Comparative Analysis of Regex and LLM-Based Methods","authors":"Neupane, B., Ku, C., & Lim, K.","venue":"International Conference on Convergent and Smart Systems (ICCSS 2025)","place":"Nairobi, Kenya","published":"June 2025","pages":null,"doi":null,"url":null,"award":"Best Paper Award","awardUrl":"/awards/best-paper-iccss-2025.pdf"},"html":"A comparison of four detection approaches: regular expressions, a closed source model, an open source model, and a fine tuned version of that open source model. The fine tuned model matched the closed source one while running locally.
"}}]}}}