Abstract :

The Swift Development Of Artificial Intelligence Has Significantly Changed The Field Of Cybersecurity. Artificial Intelligence Is Now Widely Used For Attacking And Defensive Purposes, From The Generation Of Malware, Automation Of The Attack Development, Creating New Malware Variants Which May Not Match Existing Signatures, To Malware Detection, Incident Response, Triaging, And Log Investigation. This Creates A Major Challenge For The Traditional Detection Techniques, Especially When The Malware Is Generated Or Altered Using Artificial Intelligence. Previous Research Papers Have Shown That Large Language Models (LLMs) Can Aid In Determining The AI-generated Malware. However, Many Of These Studies Were Performed Under A Controlled Environment Where The Malware Samples Were Easily Readable And Were Slightly Modified. Such Conditions Do Not Fully Represent The Real-world Scenarios, Where The Malicious Code Undergoes Transformations Such As Obfuscation, Encoding, And Encryption Before Analysis. Therefore, This Research Focuses On Assessing How Large Language Models Respond To AI-generated Malware Samples Under Different Transformation Conditions. In This Study, Various Large Language Models (LLMs) Were Examined From A Purple-team Perspective, Where Both The Offensive Sample Preparation And The Defensive Side Classification Were Evaluated Within A Safe And Controlled Environment. Malware-like And Dual-use Security Payloads Were Tested Using Structured Prompts For Analysis. Unlike Idealized Testing Conditions Where The Source Code Is Directly Readable, This Study Assesses The Samples Under Various Transformation States Such As Plain Code, Obfuscated Code, Base-92 Encoded Code, And AES-encrypted Code. The Responses Are Analyzed Using The Qualitative And Quantitative Criteria. The Findings Reveal That Models Do Provide Useful Security Insights For Malware Analysis. However, Their Detection Dependability Varies Greatly When The Original Code Is Transformed, Hidden, Or Hard To Interpret. The Study Highlights The Need For Responsible Restrictions On Malicious Prompts Usage While Focusing On The Importance Of Evaluating The AI-based Malware Detection System. This Facilitates The Practical Evaluation Of AI-assisted Malware Analysis In The Field Of Cybersecurity.

Published:

07-6-2026

Issue:

Vol. 26 No. 6 (2026)

Page Nos:

432-441

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite