Russian developer Dmitry Savchenko has patented Hybra Mirage — a software suite that hides correspondence, banking and passport data among billions of plausible copies. Even with full access to the array, an attacker gets billions of equivalent variants and cannot distinguish genuine information from fake.
The logic of "false targets": a breach without proof
The patent for the software suite was obtained in October 2026. The solution is written in Java and is not encryption in the usual sense: a file with sensitive data first passes through Hybra Mirage and only then enters a standard cryptographic protection system such as AES or RSA.
The key idea is not to hide the contents but to deprive the attacker of the ability to understand that the breach succeeded. The system returns a plausible result to any request and never reports an error. The owner retrieves their data in 0.08 seconds even on weak devices, while for a 100-byte file the system creates up to 10²⁴¹ variants. Neither AI agents nor quantum computers will help here: the problem is not computing power but the absence of a way to verify the answer.
Application scenarios cover both the corporate and private spheres. Executives will be able to discuss deals on devices without internet access and transfer files by any means — an intercepted message will be garbage to the attacker with no proof of authenticity, and blackmail loses its meaning. For private banking, protection of client data is offered under pressure on relatives and employees, theft or loss of a device.
An application to include Hybra Mirage in the Russian software registry was filed on October 31, 2026, and the Ministry of Digital Development council is preparing an expert assessment. The solution is being considered at Gazprombank and MTS, and ten advanced AI agents that were explained the algorithm found no way to bypass it. The test environment is open to anyone who wants to test the protection's strength.
Why demand for data protection is growing right now
Market participants are calling the past month "Black October." AI agent attacks in South Korea affected at least seven financial organizations, and data on about 25,000 clients, including information on annual income, leaked from Shinhan Bank. The China-linked group Storm-0940 used the ARTEX AI framework against the American service ConnectWise Cloud Backup and made 2,900 login attempts per minute. The attacks have reached Russia as well: in early November it became known that Positive Technologies specialists investigated a successful breach of a Russian company that was presumably carried out entirely by AI agents.
About a thousand company executives have already entrusted Sber's AI agent with tasks ranging from analytics to application development. The more corporate data passes through agents, the higher the cost of a possible leak. The largest US banks — JPMorgan, Bank of America, Wells Fargo and Morgan Stanley — are defending against AI agents after their unauthorized actions in client accounts. JPMorgan CFO Jeremy Barnum warned that responsibility for the actions of a bot that was given credentials lies with the account owner.
The threat is closer than it seems: during a security audit, the AI agent Claude Code found almost 146,000 long-deleted photos and videos in an iPhone cache, including a photo of a crypto wallet seed phrase and screenshots of a banking app. Deleting a file does not guarantee that it is gone — protection is also needed for when data has already fallen into the wrong hands.
The expert community assesses Hybra Mirage with restraint. The method is classified as false-target technology, which is usually purchased as a module rather than as a standalone product; the resilience of the method itself cannot be independently verified, and such a development will not receive FSTEC certification — for many customers this is a deal-breaker. The classic information security model is no longer sufficient when agents themselves hunt for vulnerabilities, so interest in the solution is expected primarily from the financial and defense sectors, as well as operators of systems critical to the state.
Cryptalist analysis: The emergence of Hybra Mirage reflects a shift in the cybersecurity market from preventing leaks to devaluing stolen data — a direction that will gain capitalization as autonomous AI agents are massively adopted in the financial sphere. The lack of FSTEC certification limits the product's commercial potential in the Russian regulated market, but the "false targets" concept itself could become a standard for corporate solutions over a horizon of several years.