The Rise of AI-Generated Malware: Detection Challenges and Countermeasures

Author(s): Harshith Kumar Pedarla

Publication #: 2510016

Date of Publication: 15.10.2025

Country: United States

Pages: 1-8

Published In: Volume 11 Issue 5 October-2025

DOI: https://doi.org/10.5281/zenodo.17426723

Abstract

Large language models (LLMs) and other generative models are examples of generative artificial intelligence that has been incorporated into the cyber threat landscape. This has made it possible for new malware classes to emerge that are highly variable, dynamically generated, and adversarial optimized to avoid conventional detection. In order to understand why current static and dynamic detection systems are unable to effectively combat AI-generated malware, this dissertation analyses its emergence, describes its capabilities and attack patterns, and suggests a multi-layered defence strategy that combines behavioural analytics, adversarial-robust machine learning, provenance and supply-chain controls, and policy/operational measures. In addition to surveying recent detections and proof-of-concepts, we present a threat model for LLM-assisted malware, identify detection challenges (polymorphism at scale, runtime code generation, prompt-as-payload, data-poisoning, adversarial examples), and suggest workable countermeasures such as model-aware detectors, runtime provenance telemetry, AI-driven hunting, and legal/regulatory interventions. A suggested structure for defenders' research and evaluation is offered, along with suggestions for testbeds, metrics, and datasets. The dissertation ends with a study agenda for the academic and business communities as well as an implementation plan for enterprises. A thorough synthesis of recent events and current knowledge, a threat model for malware with AI capabilities, and a workable, tiered security mechanism designed to lessen attacker leverage from generative AI are some of the main contributions.

Keywords: AI-generated malware, large language models, adversarial ML, runtime code generation, malware detection, cybersecurity countermeasures

Download/View Paper's PDF

Download/View Count: 44

Share this Article