Microsoft Unveils AI Agent ‘Project Ire’ for Autonomous Malware Detection

Microsoft Unveils Project Ire: Autonomous AI System Achieves 90% Accuracy in Large-Scale Malware Detection and Software Classification

  • Microsoft has developed an autonomous AI system, called Project Ire, to classify software and detect Malware without human help.
  • The system uses large language models and specialized tools to reverse engineer and analyze software files.
  • Testing shows Project Ire accurately identified 90% of software samples and kept false positives to as low as 2–4%.
  • Microsoft plans to use the prototype within its Defender organization as Binary Analyzer for enhanced threat detection.
  • The company awarded $17 million in bug bounties to security researchers worldwide in 2024–2025, with the highest single payout reaching $200,000.

Microsoft announced on August 6, 2025, the development of an autonomous Artificial Intelligence agent designed to analyze and classify software, aiming to improve large-scale malware detection. The prototype system, known as Project Ire, uses large language models and advanced reverse engineering tools to decide whether software is safe or malicious.

- Advertisement -

The company stated that Project Ire automates the complete reverse engineering process, even when there are no clues about a file’s purpose or origin. It reviews outputs from decompilers and other tools, then classifies each file. According to Microsoft, this approach cuts down on manual checks by security analysts and speeds up threat response.

Project Ire uses an API to trigger a range of tools, including Microsoft’s own memory analysis sandboxes from Project Freta, popular open-source frameworks like Ghidra and angr, and various documentation search tools. The process includes file type identification, reconstruction of control flow graphs, and validation of the chosen verdict.

The system keeps a detailed “chain of evidence” log, allowing security teams to review and adjust results as needed. In Microsoft’s tests on public Windows driver datasets, Project Ire correctly flagged 90% of samples, with a false positive rate as low as 2%. On another set of nearly 4,000 challenging files, it correctly classified almost 90% of threats, keeping false positives at 4%. “Based on these early successes, the Project Ire prototype will be leveraged inside Microsoft’s Defender organization as Binary Analyzer for threat detection and software classification,” the company stated.

Microsoft also reported it awarded $17 million in bug bounty rewards to 344 security researchers from 59 countries through its vulnerability reporting program between July 2024 and June 2025. The highest single award reached $200,000. Last year’s program awarded $16.6 million to 343 researchers from 55 countries.

- Advertisement -

The company noted its goal is to scale Project Ire to classify files from any source quickly and accurately, aiming to detect new malware in memory at large scale.

✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.

Previous Articles:

- Advertisement -

Latest News

Solana ETF Approval Seen as Imminent After S-1 Amendment Filing

The U.S. Securities and Exchange Commission (SEC) has made the 19b-4 review timeline irrelevant...

Wisconsin Bill Proposes Crypto Mining, Staking License Exemptions

Wisconsin lawmakers propose a bill to exempt individuals and businesses from money transmitter licenses...

Dormant Bitcoin Whale Moves $44M After 12 Years, Spooking Market

An inactive Bitcoin wallet holding 400 BTC, valued at over $44 million, moved funds...

Nvidia Stock Jumps as $100B OpenAI Investment Talks Fuel Optimism

NVIDIA stock rose 2% after recent losses, as several other AI-related stocks also saw...

Turkey Plans Law Allowing Watchdog to Freeze Crypto Accounts

Turkey plans to give its financial watchdog, Masak, power to freeze both bank and...
- Advertisement -

Must Read

This is How to Buy and Sell Bitcoin

Now more than ever, there are a variety of ways to enter and exit the crypto market. While this is good, the availability of...