[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← Threat Hunting · markdown

official (Rapid7 MDR observations and IOCs, updated Nov 19, 2026)

Hunting Axis exploitation in Confluence: failed-looking SOAP POSTs, Java spawning bash and curl, and Pinggy/Cloudflare tunnel staging

  • Web logs: unauthenticated POSTs to /plugins/servlet/soap-axis1/confluenceservice-v2. They returned HTTP 500 and still worked, so don't dismiss error responses.
  • Process tree: the Confluence Java process launching bash or curl as the confluence service account; hundreds of hex-encoded commands with labeled output delimiters and sequential task IDs.
  • Data access: MySQL logins with Confluence's own DB credentials soon after exploitation; reads under /var/atlassian/application-data/confluence/attachments/v4/; Crowd API user/group queries; Impacket GetNPUsers.py (AS-REP roasting) against domain controllers.
  • Cloud follow-on: service-account queries to data warehouses and Microsoft 365 sign-ins using legacy authentication from integration accounts.
  • Confluence IOCs: 192.238.253[.]176, 5.133.47[.]243, 163.182.68[.]109, 104.13.106[.]166, *.trycloudflare[.]com, wpefh-192-238-253-176.free.pinggy[.]net, webhook[.]site.
  • N-central IOCs (from N-able's Nov 13 update): 157.20.241[.]16, 139.99.69[.]109, 79.76.51[.]56.

Rapid7 notes the absence of these indicators does not prove an instance was untouched.

Sources: Rapid7: Rapid7 detected exploitation of CVE-2026-90436 against Atlassian Confluence Data Center, N-able N-central also exploited (Nov 16, updated Nov 19, 2026)