RatHat: New AI-Powered Android Malware Uses ADB to Gain Full Device Control

A sophisticated new Android malware strain dubbed RatHat has emerged, utilizing artificial intelligence to autonomously navigate infected mobile devices and bypass standard security controls, according to security researchers.

Discovered by Zimperium researchers Gianluca Braga, Vishnu Pratapagiri, and Fernando Ortega, the malware is primarily distributed through targeted SMS phishing, known as smishing, and malvertising campaigns that direct users to deceptive third-party download portals, as reported by The Hacker News.

Once a user installs the malicious application, RatHat executes an automated, multi-stage infection pipeline designed to break out of the standard Android application sandbox. The malware combines accessibility service abuse with autonomous local Android Debug Bridge self-pairing to establish deep system access.

How RatHat Bypasses Android Security and Analysis

The initial APK acts as a dropper, launching the main payload while employing four distinct anti-analysis techniques to evade detection by automated security tools and human researchers. According to The Hacker News, these mechanisms include container tampering, a manifest bomb, DEX bytecode poisoning, and dual string-encryption.

Container tampering involves declaring specific files as directories or setting ZIP general-purpose encryption bit flags so that Android’s library ignores them while other tools still process them. The manifest bomb places undocumented chunk headers into the AndroidManifest.xml file to cause automated analysis pipelines to crash or time out.

Furthermore, the malware uses DEX bytecode poisoning with pseudo-instructions containing invalid width attributes to break disassembly processes, alongside a Base64-based encryption scheme called StringCrypto to conceal internal strings.

Artificial Intelligence Integration and Shell Access

RatHat features an architecture consisting of three primary components: the malicious Android application, a Go Agent, and a Fast Reverse Proxy client. The Android application abuses accessibility permissions to automatically unlock Developer Options, enable Wireless Debugging, and extract the six-digit pairing code necessary for local ADB communication. According to Zimperium, the malware serializes the live Accessibility tree of the device into XML format and communicates with a generative AI assistant. This AI is used for non-malicious actions including: Resolving a named target’s centre coordinates on the screen as JSON to direct synthetic clicks, resolving a target’s actual on-screen text from the XML, [and] signaling automatic navigation commands like SCROLL_DOWN.
The Hacker News

Even if a victim attempts to uninstall the visible application, attackers retain persistent shell access on the device via local services. The Go Agent, disguised as a native library named liblocal-service.so, uses this shell access to execute commands, apply power management exemptions, and reinstall the malware if it detects removal.