Published on November 2, 2023, 10:19 am
Generative AI, a technology that uses artificial intelligence (AI) to create new content, is making its way into every industry, including cybersecurity. This development raises concerns among sysadmins about the potential for AI-powered malware and autonomous attacks. However, a new security company called Wraithwatch aims to counter these threats by utilizing good AI against the bad ones.
Wraithwatch was founded by Nik Seetharaman, Grace Clemente, and Carlos Más, who have firsthand experience with the constant barrage of threats faced by companies in sectors such as aerospace, defense, and finance. They believe that there is not enough dialogue about the implications of generative AI on the offensive side of cybersecurity.
Traditionally, software development involves creating code manually. However, with generative AI, developers can automate the process of creating similar code in different languages or even develop variants to improve performance or evade detection. Malware developers can exploit this capability to create multiple versions of malicious software quickly and hide from surface-level detection methods used to identify malware.
Seetharaman explains that a foreign power could easily direct an AI model to mutate itself into thousands of versions of malware and launch them simultaneously. He emphasizes the need for organizations to proactively explore the dark side of generative AI and map these techniques to their defensive strategies.
Wraithwatch is developing a cybersecurity platform that goes beyond traditional reactive methods. Rather than relying solely on manual analysis and response policies which may struggle to keep up with evolving attacks, their platform leverages AI to tailor its detection and countermeasures based on expected offensive tactics.
Current language models have limitations in their capabilities and alignment. Security researchers have already demonstrated how mainstream code-generation APIs can be tricked into aiding malicious actors. Furthermore, it’s possible for AI agents trained to interact with various platforms as human users to act as semi-autonomous threats and coordinate persistent attacks.
The solution proposed by Wraithwatch involves sitting as a command and control layer on top of existing cybersecurity tools, transforming them into a more comprehensive system. By simulating attacks using the same methods as attackers in a sandboxed environment, Wraithwatch aims to predict and characterize the variations and attacks that AI-infused malware could deploy.
The team recognizes that it’s still early days for their project, but they believe that small agile startups can outperform established companies when it comes to addressing emerging threats. With an $8 million seed round led by Founders Fund, Wraithwatch plans to have a resilient minimum viable product (MVP) deployed to design partners by Q1 of next year, with a commercial product expected by the end of 2024.
While the idea of AI agents engaging in secret cyber warfare may seem over-the-top, it’s crucial to take preventative action in this unpredictable and fast-moving landscape. Investors clearly recognize the need for solutions like Wraithwatch that can protect sensitive information from potential AI-driven attacks. As generative AI continues to advance, preparations must be made now to ensure robust cybersecurity defenses in the future.