Skip to content
NewA calmer, faster ZeroHour

Phantom53 ClickFix campaign built with Claude Code uses DNS C2

Original title: Hackers Use Claude Code to Develop Phantom53 Malware for ClickFix Attacks Targeting Global Organizations

GBHackersby Mayura Kathir ()4 min read1
Part of a story covered by 3 sourcesClickFix campaign still delivers Phantom53 and Havoc via XaocMailMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Nearly 1,000 lure pages target Brazil, Peru, India, Indonesia and Spain, and operators can load Havoc in memory.

SOCRadar says a Russian-speaking actor used Anthropic’s Claude Code to build the XaocMail phishing platform and Phantom53 malware for ClickFix attacks. Victims paste a Run-dialog command that installs an MSI, a Python runtime and scheduled tasks; Phantom53 hides commands in DNS AAAA answers and can load the Havoc Demon reflectively. The operation, prepared from October 2025 and launched in August 2026, remains active. STRU has not confirmed intent.

Unconfirmed: SOCRadar cites codebase evidence of Claude-assisted development but has not confirmed the attackers’ intent.

Key facts
  • Each victim gets a unique link to one of ten MSI payloads in Cloudflare KV.
  • Tasks imitating Edge, OneDrive, Defender and Adobe updaters beacon every 12 minutes.
  • A week-keyed SHA256 DGA produces .xyz fallback domains checked with Ed25519.
  • Aliases steal password-manager, VPN, AI-tool, browser, SSH, wallet and chat data.
  • Exfiltration goes to 145.63.138.67:8125; HavoX sits on port 8000 of that host.
Security assessment · AI, from the article

Impact: Steals credentials, secrets and wallets; loads Havoc in memory

What to do
  1. 1.Restrict the Run dialog and block unsigned MSI execution from remote URLs.
  2. 2.Watch for pythonw.exe from user-writable folders and fake updater scheduled tasks.
  3. 3.Alert on workers.dev pages and high-entropy base36 queries to public DoH resolvers.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

Indicators of compromise

auto-extracted · verify before use · export allAll
TypeIndicatorContext
ipv4145.63.138.67ainst the local Administrator account. Exfiltration goes to 145[.]63[.]138[.]67:8125, and the HavoX login panel sits on port 8000 of
ipv4162.252.198.204FastAPI Backend (HTTP 5000), DNS Server (UDP 53) IP Address 162[.]252[.]198[.]204 OPS Server (SSH:2222) IP Address 162[.]252[.]199[.]92
ipv4162.252.199.92ress 162[.]252[.]198[.]204 OPS Server (SSH:2222) IP Address 162[.]252[.]199[.]92 OPS Server (Ligolo-ng) IP Address 94[.]156[.]152[.]7
ipv4185.224.132.148speed up deployment. XaocMail is a self-hosted platform at 185[.]224[.]132[.]148:8585. It creates personalized document-sharing lures,
ipv494.156.152.7ress 162[.]252[.]199[.]92 OPS Server (Ligolo-ng) IP Address 94[.]156[.]152[.]7 Operator VPS Note: IP addresses and domains are inten

From the original article

4 min read at gbhackers.com

A Russian-speaking threat actor has built a phishing and command-and-control toolkit with help from Anthropic’s Claude Code. Per STRU, threat actors are increasingly using AI to harden infrastructure and speed up deployment. XaocMail is a self-hosted platform at 185[.]

GBHackers

By Mayura Kathir · gbhackers.com

Read the full article at GBHackers

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.