Microsoft's novel prototype autonomous artificial intelligence agent Project Ire has shown promise in identifying malware within software files, reports SecurityWeek.
With decompilers, a multi-level reasoning-capable architecture, and a tool-use API, Project Ire which was developed by Microsoft Defender Research, Microsoft Research, and Microsoft Discovery & Quantum teams yielded 90% accuracy in separating malicious and benign software, with only 2% of benign software determined to be dangerous, according to Microsoft. While another test involving nearly 4,000 files showed Project Ire to have a 4% false positive rate, the AI agent was only able to determine nearly 25% of all available malware. "Our goal is to scale the system's speed and accuracy so that it can correctly classify files from any source, even on first encounter. Ultimately, our vision is to detect novel malware directly in memory, at scale," said Microsoft, which intends to integrate Project Ire into Defender as a binary threat detection and software classification analyzer.
With decompilers, a multi-level reasoning-capable architecture, and a tool-use API, Project Ire which was developed by Microsoft Defender Research, Microsoft Research, and Microsoft Discovery & Quantum teams yielded 90% accuracy in separating malicious and benign software, with only 2% of benign software determined to be dangerous, according to Microsoft. While another test involving nearly 4,000 files showed Project Ire to have a 4% false positive rate, the AI agent was only able to determine nearly 25% of all available malware. "Our goal is to scale the system's speed and accuracy so that it can correctly classify files from any source, even on first encounter. Ultimately, our vision is to detect novel malware directly in memory, at scale," said Microsoft, which intends to integrate Project Ire into Defender as a binary threat detection and software classification analyzer.
