Elastic unknown experimental kql

Cisco Secure FMC Authentication Bypass HTTP Exploit Artifacts

Hunts for HTTP artifacts associated with exploitation of CVE-2026-20079 against Cisco Secure Firewall Management Center (FMC). The queries identify remote use of the boot-created `csm_processes` web session, a `validateLicense` request that writes attacker-controlled script content, and an `upgradeReadinessCall` request referencing `license.tmp`. These artifacts are based on independently reproduced public exploit mechanics and should be investigated as attempted or successful exploitation of the FMC management interface.

View Source

Detection Logic

FROM logs-network_traffic.http-* METADATA _source, _id, _index
| WHERE @timestamp > NOW() - 30 days
| EVAL Esql.raw_request = COALESCE(
    JSON_EXTRACT(_source, "network_traffic.http.request.original"),
    JSON_EXTRACT(_source, "request")
  )
| WHERE url.path IN ("/help/about.cgi", "/login.cgi", "/ui/user/general", "/sajaxintf.cgi", "/pjb.cgi")
    
AND Esql.raw_request LIKE "*CGISESSID=csm_processes*"
| KEEP
    @timestamp,
    source.ip,
    source.port,
    destination.ip,
    destination.port,
    http.request.method,
    url.path,
    url.query,
    http.response.status_code,
    user_agent.original,
    Esql.raw_request,
    _index,
    _id
| SORT @timestamp DESC
| LIMIT 1000



FROM logs-network_traffic.http-*, logs-modsecurity.auditlog-* METADATA _id, _index
| WHERE @timestamp > NOW() - 30 days
| WHERE http.request.method == "POST"
    
AND url.path == "/sajaxintf.cgi"
    
AND http.request.body.content LIKE "*validateLicense*"
    
AND (
        http.request.body.content LIKE "*This script was generated using Makeself*"
        
OR http.request.body.content LIKE "*#!/bin/sh*"
    )
| KEEP
    @timestamp,
    source.ip,
    source.port,
    destination.ip,
    destination.port,
    http.request.method,
    url.path,
    url.query,
    http.request.mime_type,
    http.response.status_code,
    user_agent.original,
    http.request.body.content,
    _index,
    _id
| SORT @timestamp DESC
| LIMIT 1000



FROM logs-network_traffic.http-*, logs-modsecurity.auditlog-* METADATA _id, _index
| WHERE @timestamp > NOW() - 30 days
| WHERE http.request.method == "POST"
    
AND url.path == "/pjb.cgi"
    
AND http.request.body.content LIKE "*upgradeReadinessCall*"
    
AND http.request.body.content LIKE "*license.tmp*"
| KEEP
    @timestamp,
    source.ip,
    source.port,
    destination.ip,
    destination.port,
    http.request.method,
    url.path,
    url.query,
    http.request.mime_type,
    http.response.status_code,
    user_agent.original,
    http.request.body.content,
    _index,
    _id
| SORT @timestamp DESC
| LIMIT 1000

False Positives

  • ⚠ The first query hunts Network Packet Capture events whose exported raw HTTP request contains `CGISESSID=csm_processes`. It reads both the Network Packet Capture 2.0.0 and later field layout and the legacy field layout. A request to `/help/about.cgi` may be vulnerability probing, while the cookie on `/login.cgi`, `/sajaxintf.cgi`, or `/pjb.cgi` is progressively stronger exploitation evidence.
  • ⚠ The first query requires Elastic Stack 9.4.0 or later because `JSON_EXTRACT` is unavailable on 8.19.
  • ⚠ The second and third queries hunt the file-write and privileged-execution stages in Network Packet Capture or ModSecurity audit logs. `validateLicense` can occur during legitimate license workflows, but its combination with a Makeself or shell-script marker is highly suspicious. `upgradeReadinessCall` together with `license.tmp` is the highest-confidence network artifact.
  • ⚠ Cisco FMC normally exposes its management interface over HTTPS. Elastic Network Packet Capture does not decrypt TLS; it must observe plaintext traffic after an approved TLS termination or inspection point. Merely adding TCP port 443 to the HTTP protocol configuration does not make encrypted request content visible.
  • ⚠ For Network Packet Capture, enable the HTTP protocol on the post-decryption traffic port. Enable `send_request` so the raw request and Cookie header are searchable, configure `send_headers` to include `Cookie` or use `send_all_headers`, and configure `include_body_for` for the request Content-Types used by the FMC transactions so `http.request.body.content` is populated.
  • ⚠ HTTP headers, cookies, and bodies can contain credentials, session identifiers, configuration data, and other secrets. Restrict capture to approved FMC management destinations, minimize retained fields and retention, apply role-based access controls, and follow organizational privacy and TLS-inspection requirements.
  • ⚠ The ModSecurity integration can populate `http.request.body.content`, `http.request.method`, and `url.path` from JSON audit logs when a ModSecurity reverse proxy or WAF terminates or inspects FMC HTTPS and body logging is enabled. Placing a proxy or WAF in front of FMC is not a standard deployment and should only be done when supported and approved.
  • ⚠ FMC audit syslog, ordinary flow logs, and TLS handshake metadata cannot expose these cookies, paths, or request bodies. Cisco FTD and Suricata can instead emit IDS signature alerts when they inspect the traffic; those alerts are already eligible for generic external-alert promotion.
  • ⚠ Before operational use, replay benign HTTP fixtures through an isolated capture pipeline to confirm field names, URL decoding, JSON escaping, body truncation, and Cookie retention. Do not test the authentication bypass or execute the documented exploit against production FMC systems.
  • ⚠ A high-confidence match warrants preserving FMC `/var/log/messages*`, web and audit logs, reviewing the Cisco-published `package_info.pl /var/tmp/license.tmp --lsm` IOC, restricting FMC management access, and contacting Cisco TAC. A lack of matches is not evidence that the device is patched or uncompromised.

Field Validations

Loading…

Comments (0)

Loading comments...