• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
Technology
Report

Canto Incognito campaign reportedly compromised over 3,400 servers

A threat-intelligence account says PoeLLM conceals its command-and-control address in a poem hosted on GitHub.

nolimitNO
A.HelmleA.
2 Sources, 1h ago, first seen 1h ago

TLDR

A threat-intelligence account says PoeLLM hides its command-and-control address in a public GitHub poem. Another post says hackers compromised over 3,400 servers in the Canto Incognito campaign. Citing researchers, it says the botnet is used for cryptomining, network scanning and attacks on AI infrastructure.

Combined views

42

2 Sources, first seen 1h ago

1 likes1 comments

Combined views

42

2 Sources, first seen 1h ago

1 likes1 comments

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Featured Source

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

2 Sources

nolimit@n0limitSOC🔴 THREAT INTELLIGENCE PoeLLM's Poetic C2: Hunting AI-Targeting Malware Hiding in Plain Sight on GitHub Oct 8, 2026 · 7 min read ──────────────────────────────────────── ## PoeLLM's Poetic C2: Hunting AI-Targeting Malware Hiding in Plain Sight on GitHub The call came in at 2 AM. A mid-sized tech company, flush with recent VC funding, was seeing anomalous outbound traffic from their shiny new AI inference clusters. Nothing overtly malicious flagged by their perimeter defenses—no known C2 domains, no suspicious IP ranges. Just a steady, low-level hum of data leaving their network, mostly to what looked like cloud providers and GitHub. Their junior SOC analyst was stumped. The traffic wasn't hitting any of their traditional tripwires, but the sheer volume and persistence from systems that should have been largely internal-facing set off alarm bells for them. That’s when they called us. We dove in, and it didn't take long to realize this wasn't a standard cryptominer. The indicators were too subtle, too cleverly disguised. The C2 wasn't a hardcoded IP or a DNS record in some shady TLD. It was something far more insidious: a poem, hosted on a public GitHub repository. A piece of seemingly innocuous, if somewhat amateur, verse. This was our introduction to PoeLLM. ## The Problem This isn't just about a new piece of malware; it's about a fundamental shift in attacker tradecraft. The PoeLLM campaign, dubbed "Canto Incognito" by researchers, isn't unique because it targets AI/LLM infrastructure—we've been seeing that uptick for a while now, especially with the explosion of interest in large language models. What makes it stand out is its command-and-control (C2) mechanism. The C2 address, the very heart of the botnet's ability to receive instructions and exfiltrate data, was embedded within a poem on GitHub. Think about that for a second. How many organizations are actively scanning public code repositories for poetic stanzas that might contain encoded IP addresses or domain names? How many SIEM rules are looking for beaconing to GitHub, a platform used by nearly every developer on the planet for legitimate purposes? Traditional network-based indicators of compromise (IOCs)—IP addresses, domains, hashes—are becoming less effective when attackers are leveraging legitimate services and creative obfuscation techniques. We recently saw a similar pattern with ASOS, where attackers compromised a third-party communication platform to send rogue notifications, not directly malicious traffic, but a clear abuse of a trusted channel. The data doesn't lie. Our incident response teams are seeing a steady decline in the efficacy of purely signature-based or reputation-based detection for C2. Attackers are moving away from easily blockable infrastructure to blend in with normal traffic. Supply chain attacks, like the malicious npm packages delivering Overlord RAT that were downloaded over 40,000 times, demonstrate this perfectly. They hide in plain sight, nestled within the trusted ecosystem of open-source development. This isn't just about avoiding detection; it's about minimizing the noise. A connection to GitHub is far less likely to trigger an alert than a connection to a known malicious IP in a sanctioned country. ## Why It's Getting Worse The rise of AI and automation isn't just a boon for defenders; it's an accelerator for attackers. CISA is even considering a 72-hour remediation deadline for critical flaws, a stark recognition that AI models are enabling attackers to discover and exploit vulnerabilities at an unprecedented speed. This isn't theoretical; we're seeing it in the wild. AI can be used to generate endless permutations of C2 mechanisms, to identify weak points in code for embedding payloads, and to rapidly craft spear-phishing campaigns tha... ──────────────────────────────────────── 📰 Full analysis on The Signal: https://n0limit.com/blog.html#poellm-poetic-c2-ai-malware-github-hunt #cybersecurity #threatintel #infosec1h
A.Helmle@Silencio_OffHackers have compromised over 3,400 servers in a campaign known as Canto Incognito The PoeLLM malware hides its command infrastructure inside a poem hosted on GitHub Researchers say the botnet is being used for cryptomining, network scanning and attacks on AI infrastructure1h
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    2 Sources

    nolimit@n0limitSOC🔴 THREAT INTELLIGENCE PoeLLM's Poetic C2: Hunting AI-Targeting Malware Hiding in Plain Sight on GitHub Oct 8, 2026 · 7 min read ──────────────────────────────────────── ## PoeLLM's Poetic C2: Hunting AI-Targeting Malware Hiding in Plain Sight on GitHub The call came in at 2 AM. A mid-sized tech company, flush with recent VC funding, was seeing anomalous outbound traffic from their shiny new AI inference clusters. Nothing overtly malicious flagged by their perimeter defenses—no known C2 domains, no suspicious IP ranges. Just a steady, low-level hum of data leaving their network, mostly to what looked like cloud providers and GitHub. Their junior SOC analyst was stumped. The traffic wasn't hitting any of their traditional tripwires, but the sheer volume and persistence from systems that should have been largely internal-facing set off alarm bells for them. That’s when they called us. We dove in, and it didn't take long to realize this wasn't a standard cryptominer. The indicators were too subtle, too cleverly disguised. The C2 wasn't a hardcoded IP or a DNS record in some shady TLD. It was something far more insidious: a poem, hosted on a public GitHub repository. A piece of seemingly innocuous, if somewhat amateur, verse. This was our introduction to PoeLLM. ## The Problem This isn't just about a new piece of malware; it's about a fundamental shift in attacker tradecraft. The PoeLLM campaign, dubbed "Canto Incognito" by researchers, isn't unique because it targets AI/LLM infrastructure—we've been seeing that uptick for a while now, especially with the explosion of interest in large language models. What makes it stand out is its command-and-control (C2) mechanism. The C2 address, the very heart of the botnet's ability to receive instructions and exfiltrate data, was embedded within a poem on GitHub. Think about that for a second. How many organizations are actively scanning public code repositories for poetic stanzas that might contain encoded IP addresses or domain names? How many SIEM rules are looking for beaconing to GitHub, a platform used by nearly every developer on the planet for legitimate purposes? Traditional network-based indicators of compromise (IOCs)—IP addresses, domains, hashes—are becoming less effective when attackers are leveraging legitimate services and creative obfuscation techniques. We recently saw a similar pattern with ASOS, where attackers compromised a third-party communication platform to send rogue notifications, not directly malicious traffic, but a clear abuse of a trusted channel. The data doesn't lie. Our incident response teams are seeing a steady decline in the efficacy of purely signature-based or reputation-based detection for C2. Attackers are moving away from easily blockable infrastructure to blend in with normal traffic. Supply chain attacks, like the malicious npm packages delivering Overlord RAT that were downloaded over 40,000 times, demonstrate this perfectly. They hide in plain sight, nestled within the trusted ecosystem of open-source development. This isn't just about avoiding detection; it's about minimizing the noise. A connection to GitHub is far less likely to trigger an alert than a connection to a known malicious IP in a sanctioned country. ## Why It's Getting Worse The rise of AI and automation isn't just a boon for defenders; it's an accelerator for attackers. CISA is even considering a 72-hour remediation deadline for critical flaws, a stark recognition that AI models are enabling attackers to discover and exploit vulnerabilities at an unprecedented speed. This isn't theoretical; we're seeing it in the wild. AI can be used to generate endless permutations of C2 mechanisms, to identify weak points in code for embedding payloads, and to rapidly craft spear-phishing campaigns tha... ──────────────────────────────────────── 📰 Full analysis on The Signal: https://n0limit.com/blog.html#poellm-poetic-c2-ai-malware-github-hunt #cybersecurity #threatintel #infosec1h
    A.Helmle@Silencio_OffHackers have compromised over 3,400 servers in a campaign known as Canto Incognito The PoeLLM malware hides its command infrastructure inside a poem hosted on GitHub Researchers say the botnet is being used for cryptomining, network scanning and attacks on AI infrastructure1h
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet