Two Days After a Patch, Red Heron Scanned 1,386 Gitea Servers
Security / news
Two Days After a Patch, Red Heron Scanned 1,386 Gitea Servers
Acronis says the Chinese-speaking group turned a fixed Gitea flaw into a rootkit-based campaign that compromised organizations in five countries within a week.

A Chinese-speaking hacking group scanned 1,386 internet-facing Gitea servers across seven countries within two days of a fix shipping for a critical remote-code-execution flaw, and went on to compromise at least 11 organizations, Acronis's Threat Research Unit said Sept. 13. The unit tracks the group as Red Heron.
Gitea, a self-hosted alternative to GitHub, patched the flaw, tracked as CVE-2026-60004 and rated 9.8 out of 10 on the CVSS severity scale, in version 1.27.1 on July 27. The bug lets an attacker abuse Git's three-way merge inside Gitea's diffpatch feature to write malicious files directly into a repository's hook scripts, according to Cyber Security News, which reviewed the same campaign. Those hooks then run with the privileges of the Gitea process itself.
How fast CVE-2026-60004 went from patch to exploitation
Acronis researcher Subhajeet Singha wrote that Red Heron converted public proof-of-concept code into an automated framework within two days of the advisory. Weaponization began July 29, large-scale scanning was complete by July 30, and Acronis found the group's staging server on Aug. 4, a week after the patch shipped.
| Metric | Count |
|---|---|
| Gitea versions affected | 1.17 through 1.27.0 |
| Instances scanned | 1,386 |
| Countries scanned | 7 |
| Taiwan-focused targets | 477 |
| Confirmed compromised organizations | 11 |
What Red Heron leaves behind
The group's implant, JITTERLY, is written in C++, supports more than 30 post-exploitation commands and encrypts its traffic with AES-128-GCM, according to Acronis. It pairs with a separate rootkit called SIXZUT, which loads via Linux's LD_PRELOAD mechanism and hooks 15 library functions to hide its own files, processes and network connections, and to relaunch itself if a defender kills it.

Where the campaign actually landed
Acronis confirmed victims in Canada, Argentina, Taiwan, the United States and Sri Lanka. A Canadian renewable-energy company suffered the most extensive compromise the researchers documented, with the group reaching its source code, HR and CRM systems, authentication services and Docker infrastructure. A separate Taiwan-based victim's compromised Synology NAS gave Red Heron administrative access to a Proxmox virtualization cluster.
Acronis assigns the Taiwan-heavy targeting, the group's use of Simplified Chinese in its operational material, and its classification of Taiwan as domestic territory to "moderate confidence" that Red Heron operates within a PRC-linked context. That is a lower bar than attribution to a specific government unit, and Acronis does not claim more than that.
The speed of the Gitea campaign echoes the automated re-infection pattern in npm's own Shai-Hulud worm, though Red Heron's tooling targets self-hosted git infrastructure rather than package registries. It also fits the pattern behind Cisco's own September disclosures: once a serious n-day flaw is public, named groups now weaponize it in days, not weeks.
Gitea has not published an updated advisory since July 28, and Acronis has not said whether it notified all 11 confirmed victims directly or through national computer emergency response teams.
Sources
More in Security
- 01WordPress 7.1.1 Closes the Click2Shell Remote Code ChainResearcher Paulos Yibelo's chain turned one clicked link into server-side code execution by tricking WordPress into silently installing a theme with no authorization checks at all.
- 02Oracle's September Update Fixes Six Perfect-10 BugsNone of the maximum-severity flaws is confirmed under attack yet, but Oracle is now shipping security patches on a monthly rather than strictly quarterly calendar.
- 03Mathspace Breach Exposes Data on 1.08 Million StudentsAttackers had 17 days between reaching Mathspace's reporting database and downloading it, and the company patched the underlying Metabase flaw only after both had happened.
- 04BragJack Hijacks AI Agents Built Into Five BrowsersA single malicious extension could seize Chrome's Gemini, Edge's Copilot, Opera Neon, Perplexity Comet and Claude in Chrome without bypassing any model's guardrails or writing a single injected prompt.