Orange Unveils AI‑Driven Maldive and Merlin Tools to Boost Malware Detection and Test Defenses

Globally, 3.2 new malware or potentially unwanted applications are discovered each second, highlighting a persistent and growing threat. This scale of digital adversary necessitates advanced defense mechanisms. Orange addresses this challenge with two new AI-driven cybersecurity tools: Maldive and Merlin.

Maldive functions as a comprehensive malware detection and analysis pipeline, leveraging static, dynamic, and AI-based analysis mechanisms to scrutinize malicious files. Its core component, MalID, provides detailed profiles on malware, including its nature and behavior. The system demonstrates a 90‑95% detection and classification accuracy rate on test sets, completing analyses in under one minute.

A key advantage of Maldive is its sovereign code, which gives Orange full control over the end-to-end pipeline. This approach avoids external dependencies, enhances data control, and supports Orange's 2026 roadmap for improving model explainability, according to Research Engineer Benjamin Marais.

Complementing Maldive, the Merlin project employs artificial intelligence to test the resilience of existing detection systems. Merlin uses reinforcement learning, where an AI program learns to modify malware structures to make them undetectable. This process effectively audits and benchmarks antivirus software and other security models, guiding future security investment decisions.

Orange is currently testing Merlin with French cybersecurity specialist HarfangLab, continuously refining its catalog of modification actions. Maldive, now in its prototype stage, shows strong potential for operational deployment, with discussions underway within the Orange Group. The cybersecurity community will observe how these AI-powered tools enhance detection capabilities and fortify defenses against evolving threats.