Description

A new wave of AI-powered "Gray Bots" has been observed aggressively targeting web applications with over 17,000 requests per hour. These bots stand out due to their ability to mimic human behavior using machine learning, enabling them to bypass traditional rate-limiting and bot detection systems. Unlike conventional brute-force or scripted attacks, Gray Bots dynamically adjust their traffic patterns, spread their requests across multiple IP addresses, rotate user agents, and introduce randomized delays, blending in with normal traffic. Their primary focus is on authentication endpoints and API gateways, probing for vulnerabilities such as credential stuffing opportunities. Discovered by Barracuda security researchers in February 2025, the bots display advanced evasion tactics and authentication bypass techniques not seen in traditional botnets. What makes them more dangerous is their ability to learn from failed or blocked attempts and adapt accordingly. These Gray Bots are causing real-world consequences, including account takeovers, service degradation, increased infrastructure strain, and data theft—especially for financial institutions and e-commerce platforms. The volume of requests also inflates operational costs for affected organizations, as defenses struggle to scale effectively under such persistent and stealthy attacks. The bots propagate through compromised websites, injecting polymorphic JavaScript code that creates temporary browser-based botnets. Rather than relying on traditional malware, the script runs within legitimate user sessions by creating temporary background web workers. This method ensures the bots persist even after users leave the compromised page. Through encrypted WebSocket connections, attackers can coordinate large-scale operations remotely, turning average website visitors into unwitting participants in distributed attacks against target platforms.