VYPR
researchPublished Nov 20, 2026· 1 source

Bazuka Android Malware Leverages AI and Fake Overlays for Sophisticated Credential Theft

A new Android malware framework, Bazuka, integrates large language models for victim management and employs deceptive screen overlays to conceal its malicious activities.

A novel Android malware framework dubbed Bazuka has emerged, distinguished by its integration of large language model (LLM) capabilities for enhanced victim management and control at scale. Discovered by mobile security firm ThreatFabric in late October 2026, Bazuka was identified during an investigation into an NFC-relay banking Trojan, signaling a significant advancement in mobile threat sophistication.

The Bazuka framework is comprehensive, encompassing a dropper/downloader, a core module, and an extensive suite of 23 specialized modules. These modules collectively enable a wide range of malicious functions, including sophisticated overlay techniques, keylogging, interception of two-factor authentication (2FA) codes and SMS messages, remote device control, and cryptocurrency address swapping.

A particularly innovative feature of Bazuka is its novel overlay mechanism. The malware can capture screenshots of the user's home or lock screen and display these as an overlay on demand. This technique effectively masks the malicious activities occurring in the background, such as credential harvesting or unauthorized actions, making it difficult for users to detect compromise.

Further enhancing its operational capabilities, Bazuka incorporates artificial intelligence (AI) within its control panel. This AI integration assists operators in identifying high-value victims and managing infections more efficiently. The framework supports both locally hosted AI models and those accessed via an API, with sample prompts suggesting the AI can score victims based on their attractiveness as targets. The control panel also facilitates the creation of accounts for 'traffers'—cybercriminals responsible for luring victims—and the construction of droppers designed to bypass Google Play's review process by carefully selecting requested permissions.

Distribution channels for Bazuka appear to be varied. Dropper applications can be disseminated through traffers using social media ads or deceptive websites, as well as potentially through the Google Play Store. One observed dropper masqueraded as a premium, adult-themed version of TikTok, using the package name "com.tiktok.premium." Attack chains observed include a dropper that requests third-party installation permissions before downloading a stager APK, or a downloader that directly fetches the main payload.

Once the core and necessary modules are loaded, Bazuka functions as a potent keylogger and overlay Trojan. It meticulously monitors user input, displaying overlays at opportune moments to conceal its operations. All actions are dictated by commands received from its command-and-control (C2) servers, communicated over two distinct WebSocket channels: one for control commands and another for data exfiltration and screen capture streams.

The core module acts as the central orchestrator, managing C2 communication, loading additional modules, dispatching commands, gathering device information, and handling permission management and component updates. It leverages Android Debug Bridge (ADB) shell commands and accessibility services to automate permission requests and grants. Notably, the core can enable Wireless Debugging, intercepting the pairing PIN to establish a secure TLS connection to the ADB daemon, and can also disable Google Play's protection features and initiate a self-destruct sequence to erase traces of its presence.

With its extensive array of 23 modules, Bazuka is equipped for comprehensive data and cryptocurrency theft. Modules facilitate call forwarding, cryptocurrency wallet address swapping, contact harvesting, device control via ADB, file system access, detailed input recording (including sensitive information like bank card numbers and passwords), location tracking, media capture, and remote control. The malware also includes functionalities for managing notifications, displaying phishing overlays, intercepting SMS messages for verification codes, and employing anti-analysis measures. Specialized modules target NFC interactions for contactless card fraud and include mechanisms to prevent uninstallation or force-stopping.

Synthesized by Vypr AI