LSU Research Bites: Malware-Detecting Large Language Model a Game-Changer for Cybersecurity

November 06, 2025

To detect, remove, and prevent malware on your devices, cyber experts have to figure out where the code came from and what it’s doing.

But dissecting malware’s code and behaviors to understand its mechanisms is becoming a find-a-needle-in-a-haystack problem. Modern malware is complex and adept at “hiding” on your devices.

To address this problem, LSU cybersecurity researchers, including Aisha Ali-Gombe and James Ghawaly in the Division of Computer Science and Engineering, have created a malware-detecting large language model called MalParse.

Problem: Malware, malicious software designed to disrupt computer systems, is increasingly complex and adept at "hiding" on devices.

Solution: LSU researchers created MalParse, a generative AI model that identifies and classifies malware and can also explain its decisions like a cybersecurity expert.

Impact: MalParse is a game-changer for faster detection and prevention of increasingly complex malware, and training future analysts.

As an LLM like ChatGPT, MalParse can not only categorize malware, but also explain why it made any given decision—unlike other “black box” AI models that make decisions based on unknown factors. It can pinpoint precise malicious code snippets.

Aisha Ali-Gombe

In testing, MalParse classified malware with 77% accuracy without prior training and generated a detailed description of the malicious behavior’s root cause.

“LLMs can act as semantic reverse engineers, capable of tracing logic chains, identifying API misuse, and explaining intent, all through language reasoning rather than traditional analysis techniques,” Ali-Gombe said. “This system replaces brittle, hand-engineered rules with intelligent, adaptive reasoning, showing that LLMs can reason about security and privacy artifacts the way experts do.”

The natural-language outputs MalParse can also be used to train the next generation of analysts, making complex malware behavior explainable in plain English.

Read the paper: Walton, B. J., Khatun, M. E., Ghawaly, J. M., & Ali-Gombe, A. (2024, December). Exploring large language models for semantic analysis and categorization of android malware. In 2024 Annual Computer Security Applications Conference Workshops (ACSAC Workshops) (pp. 248-254). IEEE.

Next Steps

Let LSU put you on a path to success! With 330+ undergraduate programs, 70 master's programs, and over 50 doctoral programs, we have a degree for you.

Experience LSU Find Your Major Become a Tiger