Root Causes
According to Check Point Research’s recent findings, two primary factors enable these attacks:
- AI Training Data Contamination: Large language models trained on database documentation and deployment tutorials now generate code containing the same vulnerable default credentials present in their training datasets. This creates a self-perpetuating cycle where AI-assisted development inadvertently propagates security weaknesses.
- Legacy Stack Vulnerabilities: Outdated web server configurations, particularly XAMPP deployments, continue operating with exposed administrative interfaces and inadequate security hardening.
Technical Evolution
Historical Context
GoBruteforcer (GoBrut) emerged in early 2023 targeting Unix-like systems across x86, x64, and ARM architectures. Initial variants deployed IRC-based command infrastructure alongside web shells and credential-scanning modules.
By September 2025, researchers discovered overlap between GoBruteforcer infrastructure and the SystemBC malware ecosystem, indicating shared botnet resources.
Recent Enhancements
Mid-2025 samples demonstrate significant technical sophistication:
- Rewritten IRC Component: Core bot functionality migrated to heavily obfuscated Golang implementation
- Enhanced Evasion: Advanced process masquerading and improved persistence mechanisms
- Dynamic Credential Management: Operators maintain rotating credential pools that refresh multiple times weekly, allowing campaigns to pivot between target sectors
Attack Methodology
Credential Selection Strategy
The malware employs curated username-password combinations derived from multiple sources:
- Generic weak credentials (myuser:Abcd@123, appeaser:admin123456)
- Cryptocurrency-specific accounts (cryptouser, appcrypto, crypto_app)
- Common CMS defaults (root, wordpress, wpuser for phpMyAdmin)
FTP attacks utilize a static credential set hardcoded within the binary, specifically targeting web hosting environments and default service accounts.
Infection Chain
- Initial Compromise: Scanning identifies internet-exposed FTP services, particularly XAMPP installations
- Web Shell Deployment: Successful authentication enables PHP web shell upload
- Bot Installation: Architecture-specific shell scripts download and execute the IRC bot payload
- Botnet Integration: Compromised hosts assume one of three operational roles:
- Credential brute-forcing nodes targeting FTP, MySQL, PostgreSQL, and phpMyAdmin services
- Payload distribution servers for lateral movement
- Backup command-and-control infrastructure ensuring operational resilience
Cryptocurrency-Focused Operations
Analysis of compromised infrastructure revealed specialized modules querying TRON blockchain addresses via tronscanapi[.]com to identify wallets containing funds, confirming deliberate targeting of blockchain-related systems.
Broader Threat Landscape
Concurrent threat activity involves systematic internet-wide scanning for misconfigured proxy servers that could facilitate unauthorized access to commercial LLM platforms.
- Campaign One (October 2025 – January 2026): Exploited server-side request forgery vulnerabilities in Ollama model pulls and Twilio SMS webhooks. Attribution evidence suggests security research or bug bounty operations.
- Campaign Two (December 28, 2025 onward): High-intensity enumeration targeting LLM API endpoints from multiple providers including Alibaba, Anthropic, DeepSeek, Google, Meta, Mistral, OpenAI, and xAI. Over eleven days, source IPs 45.88.186[.]70 and 204.76.203[.]125 generated 80,469 scanning sessions across 73+ model endpoints.
Implications
This campaign highlights the compounding security challenges created when AI-generated code propagates vulnerable defaults, legacy infrastructure remains unpatched, and automated attack tools continue advancing in sophistication. Organizations must address exposed services, enforce strong authentication, and audit AI-assisted development outputs for security weaknesses.
References
- CheckPoint Inside GoBruteforcer: AI-Generated Server Defaults, Weak Passwords, and Crypto-Focused Campaigns
- SecurityWeek GoBruteforcer Botnet Targeting Crypto, Blockchain Projects
- Paloalto Unit 42 GoBruteforcer: Golang-Based Botnet Actively Harvests Web Servers
- DarkReading Multipurpose GoBruteforcer Botnet Targets 50K+ Linux Servers