Ghost CMS CVE-2026-26980: Content API SQL injection mass-exploited to steal Admin API keys and turn 700+ sites into ClickFix lures
Attackers used a Ghost CMS SQL injection to poison 700+ sites with fake verification pages that push stealers to users
Analysis
CVE-2026-26980 is an unauthenticated SQL injection in the Ghost CMS Content API, present from 3.24.0 to 6.19.0, fixed in 6.19.1 and rated CVSS 9.4 on Ghost's advisory; because it gives access to a site's API keys, Ghost recommends reviewing staff users and rotating keys after upgrading (Ghost, GHSA-w52v-v783-gw97). There is no application-level workaround, since the Content API key is public by design; a reverse proxy or WAF rule blocking Content API requests whose filter contains slug:[ (or slug%3A%5B) is the temporary mitigation (Ghost, GHSA-w52v-v783-gw97). The fix shipped in February, but many sites did not install it (BleepingComputer, 2026-05-24).
XLab detected the campaign on 2026-05-07, confirmed 156 victim domains by 2026-05-10 and more than 700 in its enumeration of 2026-05-17, when it also found a second attack group; the two groups sometimes replace each other's code on the same site (XLab, 2026-05-21). XLab names Harvard University, Oxford University and Auburn University among the well-known sites affected, and its victim table is 48.1% personal blogs or independent sites (XLab, 2026-05-21); BleepingComputer adds DuckDuckGo and describes the mix as university portals, AI/SaaS companies, media outlets, fintech firms, security sites and personal blogs (BleepingComputer, 2026-05-24).
The attackers used the SQL injection to read each site's Admin API key, which has management permissions over articles, themes and users, then called the Admin API (PUT /ghost/api/admin/posts/:id/) to append a JavaScript loader to articles in bulk (XLab, 2026-05-21). The loader fetches a cloaking script, which XLab traces to the commercial Adspect service, that fingerprints each browser and shows only visitors it judges to be real victims a forged Cloudflare "Verify you are human" page in an iframe (XLab, 2026-05-21). While the page silently downloads an archive, it tells the visitor to press WIN+R and paste a command that moves the archive to %TEMP%, extracts it with tar and starts the batch file inside it minimized; the batch file runs hidden PowerShell that downloads a DLL and starts one of its exported functions through rundll32 (XLab, 2026-05-21). From 2026-05-16 the DLL fetched a stealer, an Electron application that persists through Electron's login-item API and polls its server every 30 seconds for JavaScript or executables to run (XLab, 2026-05-21). BleepingComputer also lists DLL loaders and JavaScript droppers among the payloads (BleepingComputer, 2026-05-24).
Cited evidence
The attacker exploited the high-risk SQL injection vulnerability CVE-2026-26980 in Ghost CMS to obtain the target site's Admin API Key without authorization
Second Round of Compromised Site Enumeration Completed: 700+ victim domains confirmed; new threat actor identified.
Has management permissions over articles, themes, users, etc.
threat actors planted malicious code on the websites of Harvard University, Oxford University, Auburn University, and DuckDuckGo
Updates2
Ghost's advisory says there is no application-level workaround and that restricting the Content API key does not mitigate the flaw, because the key is public by design; the temporary control is a reverse-proxy or WAF rule blocking slug:[ filters (Ghost, GHSA-w52v-v783-gw97). The earlier advice to restrict the Content API to trusted origins would not have protected a site.
Ghost's advisory gives the affected range, 3.24.0 to 6.19.0, without distinguishing hosted from self-hosted sites (Ghost, GHSA-w52v-v783-gw97). XLab's figure of more than 700 counts poisoned domains, confirmed in its enumeration of 2026-05-17, and XLab names Harvard University, Oxford University and Auburn University among them (XLab, 2026-05-21); BleepingComputer adds DuckDuckGo (BleepingComputer, 2026-05-24). In the chain XLab traced, a cloaking script it attributes to the commercial Adspect service shows only visitors it judges to be real victims a forged Cloudflare "Verify you are human" page in an iframe; that page itself downloads the archive in the background, and the command the visitor pastes only moves, extracts and starts it (XLab, 2026-05-21). XLab's site-side checks are PUT /ghost/api/admin/posts/:id/ requests in the Ghost logs, unknown Admin API keys and <script> tags added to posts, Code Injection settings or themes; after upgrading it advises rotating the Admin and Content API keys, the administrator password and sessions, and keeping at least 30 days of Admin API call logs (XLab, 2026-05-21).
Sources3
Revision history
- Published 2026-05-25-d675ef38
- Correction 2026-10-11T1307Z-audit
Restricting the public Content API is not a mitigation, according to Ghost's own advisory; the WAF rule on slug filters is. Ghost's advisory is dated 2026-02-16, while BleepingComputer dates the fix's release to 19 February and XLab dates public disclosure to the same day; the sources describe an unauthenticated arbitrary database read rather than a boolean-blind injection with time-delay artefacts, and the file name, the 2024 ClickFix history, the region and sector, the Ghost(Pro) statement and the self-quoted evidence had no source. The entry is rewritten from Ghost's advisory, XLab's report and BleepingComputer with the affected range, the WAF mitigation, the campaign timeline, detection and takeaway lines, verbatim evidence, a rating and an ATT&CK mapping.
Changed: title headline summary tags regions sectors techniques affected_products cves sources evidence classification body deep_dive deep_dive_category sourcing_note
- Improvement 2026-10-21T1149Z-audit
Ghost's advisory does not limit the flaw to self-hosted sites, and XLab's figure of more than 700 counts poisoned domains as of its 2026-05-17 enumeration, not sites whose keys were confirmed stolen. Added from the campaign reports: the named victims (Harvard, Oxford and Auburn per XLab, DuckDuckGo per BleepingComputer), the Adspect-derived cloaking and fake Cloudflare iframe, the background download that the pasted Run-dialog command unpacks, XLab's site-side log and key checks with its credential-rotation list, and a fuller ATT&CK mapping.
Changed: body evidence headline summary techniques
AI-generated · no human review · this permalink is the shareable record for the finding · verify operationally critical claims against the linked primary source.