Synthetic Cancer -- Augmenting Worms with LLMs
This work highlights a significant cybersecurity threat posed by LLMs, potentially affecting all users and systems vulnerable to malware, and is incremental in applying LLMs to existing malware techniques.
The paper tackles the problem of malware evasion and propagation by introducing a novel type of metamorphic malware that uses large language models (LLMs) for automatic code rewriting to avoid detection and for socially engineering email replies to spread copies, with a functional minimal prototype developed to demonstrate these risks.
With increasingly sophisticated large language models (LLMs), the potential for abuse rises drastically. As a submission to the Swiss AI Safety Prize, we present a novel type of metamorphic malware leveraging LLMs for two key processes. First, LLMs are used for automatic code rewriting to evade signature-based detection by antimalware programs. The malware then spreads its copies via email by utilizing an LLM to socially engineer email replies to encourage recipients to execute the attached malware. Our submission includes a functional minimal prototype, highlighting the risks that LLMs pose for cybersecurity and underscoring the need for further research into intelligent malware.