Many business models rely on proprietary software. Not distributing source code has for a long time been considered to safeguard a company’s intellectual property and to limit their adversaries’ ability to understand, modify, or misuse their applications. In short, compilation was seen as a form of software protection. An underlying assumption was, and is, that reverse engineering binaries is significantly harder than analyzing source code, particularly when they are protected with additional protection techniques such as, e.g., obfuscations.

The recent rise of AI-based reverse engineering strategies has challenged that assumption. Emerging methods include machine-learning-based decompilation; large language models to deobfuscate, summarize, and reason through obfuscated code; and agentic AI to automatically autonomously orchestrate advanced software analysis tools. At the same time, these methods also allow for more effective and efficient malware analysis.

The rapid evolution of AI-assisted reverse engineering necessitates an assessment of its implications for software business models, software verification practices, and the effectiveness of existing protection technologies. There is a need to evaluate the capabilities and limits of AI-based reverse engineering, and to understand how defensive techniques can adapt in response, e.g., by developing novel obfuscation strategies that still defy AI-based tools.

This Dagstuhl Seminar aims to establish a clear picture of the status and reach of this arms race by bringing together experts in (adversarial) reverse engineering, software protection, security economics, software engineering, and AI. The scope includes malware, goodware, and third-party software components. Given the speed at which AI capabilities are evolving, the seminar will provide an open platform for discussing related and emerging themes, fostering a broader conversation on the role of AI in reverse engineering and software protection.

• the status of the arms race of software protection vs. analysis tools and the trajectory of AI-based reverse engineering methods;
• the reliability of those methods, and methodologies to measure, evaluate, and compare their effectiveness and efficiency;
• their scalability limits and economic implications;
• how AI capabilities impact work flows and reverse engineering strategies, and how they can serve as decision support for software protection strategies;
• the most promising directions for preventive, anti-AI analysis protection techniques;
• educational and ethical frameworks required to respond effectively.

Creative Commons BY 4.0

Mila Dalla Preda, Bjorn De Sutter, Marion Marschalek, and Sebastian Schrittwieser