# Investigate Cisco ASA and FTD restarts

The Sigma rule and four native search starters find documented Cisco startup messages in existing appliance syslog. They help an administrator investigate restarts while responding to CVE-2026-76418. A result does not establish exploitation, and a clean search does not establish safety. Patch decisions must use the [Cisco advisory](https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-asaftd-tunnel-dos-s6mgkb75), deployed software, and configuration.

## Start with standard Sigma

Use `sigma/cisco_asa_ftd_startup.yml` with your normal Sigma workflow. The rule uses `logsource.product: cisco`, `author: Deception Check`, and standard full-event keyword matching. There is no Deception Check service, normalized field, custom table, or `schema.md` prerequisite. No new parsed fields are required.

The keyword list enumerates ASA and FTD, severity values 0 through 7, and startup IDs 199002 and 199005 with the message-ID colon retained. Informational severity 6 is the documented ordinary form. The level is informational and status experimental: planned startup is an expected match, and a hit is not a maliciousness verdict.

Select your actual ASA/FTD source in the normal converter or SIEM configuration. `product: cisco` and the definition text do not automatically create an index, identify your appliances, or configure a backend. Your backend must support full-event keyword matching against original appliance message text. If it cannot, use the documented native query for your platform below. Do not rename a custom normalization field to make this appear compatible.

Sigma string matching is case-insensitive. The supplied native regex queries retain the documented uppercase ASA/FTD prefixes. Lowercase and mixed-case fixtures are explicit Sigma positives; adjust a native search deliberately if your ingestion changes prefix case. No cross-backend identity or execution parity is claimed.

Local checks cover the rule schema, actual pySigma parsing, all 32 tokens, negative and missing-data cases, prefixes/suffixes, multiline messages, and case behavior. One actual Splunk backend conversion was inspected with a normal ASA/FTD source guard. Splunk free-text segmentation differs from raw regex matching; conversion is not proof of operational equivalence. The conversion record is internal validation material, and the native SPL regex is the distributed operational reference. No vendor or tenant query was executed.

## Start with exposure

Have the firewall administrator review the deployed configuration and Cisco's affected-release guidance. Check SSL remote access VPN reachability and secondary authentication. With `tunnel-group-list enable`, inspect all tunnel groups, including `DefaultWEBVPNGroup`; without it, inspect that default group. Do not rely on a successful login or a list of visible profiles to decide exposure. For FTD, inspect the deployed remote access VPN policy and resulting running configuration through FMC or FDM. Follow Cisco's fixed-release guidance. Treat its temporary mitigation as a security and usability change that needs administrator evaluation.

## Before searching

- Collect appliance syslog from the affected inventory into a central store. Verify informational messages, normally severity 6, are forwarded and retained. Validate facility/severity filters, timestamps, time zone, device identity, and reachability of the log collector. Do not reboot a production device merely to test this pack.
- Splunk: replace `YOUR_CISCO_SYSLOG_INDEX` in `hunts/cisco-startup.spl` with the index containing appliance syslog. The query uses `_raw` and standard event metadata, with documented `cisco:asa` and `cisco:ftd` sourcetypes. If your onboarding assigns different types, inspect real records and adjust the source selector. This is an ordinary Splunk search; it uses no Enterprise Security data model. Existing ingestion/search access and normal Splunk licensing are prerequisites.
- Microsoft Sentinel: `hunts/cisco-startup.kql` requires appliance messages in the native `Syslog` table, typically through a Linux forwarder and the Syslog via Azure Monitor Agent connector. Configure the data collection rule to retain the appliance facility and informational messages. This does not search `CommonSecurityLog`; a CEF-only deployment needs a separate mapping review. A configured Log Analytics workspace, connector permissions, and normal Azure/Sentinel ingestion and query entitlements are prerequisites. Check applicable ingestion and retention charges with your administrator.
- Sumo Logic: replace `YOUR_CISCO_SYSLOG_CATEGORY` in `hunts/cisco-startup.sumo` with the source category holding the original appliance syslog. Set the search time range to the last seven days, or your incident window, in the UI. The query uses the native `_raw` message and needs no package-specific parser. Third-party syslog ingestion, retained messages, and account search access are prerequisites; check your ingestion and retention entitlement. Inspect source-host metadata against the appliance inventory before attributing results.
- CrowdStrike Falcon LogScale: select the repository or view containing ingested Cisco appliance syslog and the last seven days, or your incident window, before running `hunts/cisco-startup.logscale`. It searches native `@rawstring`; inspect each event's original header and your documented source metadata to identify the appliance. Third-party log ingestion, retention, and search entitlement are prerequisites. This is LogScale coverage over imported firewall logs, not Falcon endpoint EDR coverage. Do not run it against an endpoint-only dataset and interpret an empty result as firewall safety.
- Verify inventory-to-record identity first. Splunk `host` can reflect forwarding configuration. Retain `HostName`, `HostIP`, `Computer`, and `CollectorHostName` in KQL results; `CollectorHostName` identifies the collector, and `HostIP` can be blank or a placeholder. Do not group distinct firewalls under one collector or infer attacker addresses from these fields.

## Read the results

The searches return individual startup records: message 199005 indicates startup beginning; 199002 indicates startup completion. Those can describe two stages of one boot, so two records are not necessarily two restarts. The regex accepts Cisco severity values 0 through 7 and requires the message-ID colon, avoiding a match on a longer numeric ID.

Compare each candidate boot with maintenance records, power or platform events, uptime, high-availability behavior, and VPN outage reports. Planned restarts, upgrades, and unrelated failures are expected matches. Use recorded device timestamps and receipt timestamps to account for forwarding delay. Preserve surrounding raw syslog and relevant diagnostic files through your established incident process.

For an unexplained restart, Cisco documents `show version` on ASA and `show version system` on FTD to examine application uptime. Follow its [unexpected reboot troubleshooting guide](https://www.cisco.com/c/en/us/support/docs/security/adaptive-security-appliance-asa-software/217663-troubleshoot-asa-or-ftd-unexpected-reloa.html) for platform-specific evidence collection and TAC escalation. Keep configurations and diagnostic bundles in approved internal storage because they can contain sensitive operational data.

Message 199001 records a user-initiated reload. It may provide context, but its absence is not evidence of an attack; do not exclude startup candidates automatically because a 199001 record appears nearby. Lost logs, buffering, clock differences, or collection gaps can obscure a sequence. An absence of startup events is not a collection-health test.

## Validation and scope

These are investigation starters, not a CVE exploit signature or production alert rules. They make no claim to identify the crafted request, distinguish malicious and accidental reboots, or observe attempts that do not produce a retained startup record. There is no endpoint-agent equivalent in this pack, and no custom Deception Check event schema is required.

Author checks used documented source formats and synthetic positive, negative, and missing-data fixtures against the exact regex strings extracted from all four query files. Python substring matching checks the SPL, KQL, and LogScale predicates; Sumo's explicit dot-all wrapper is checked with both Python whole-string and substring matching, including prefixes, suffixes, multiline messages, and trailing newlines. This checks message matching locally, not the backend regex engines or query languages. No Splunk, Kusto, Sumo, or LogScale query execution, appliance test, or customer-tenant validation was performed. Sigma schema/parsing and a Splunk conversion were checked separately as described above. Have your platform administrator run a read-only sample search and verify device identity, timestamp handling, source selection, and actual retained messages before operational use.

The portable Sigma rule uses full-event keywords rather than a normalized field mapping. The native queries remain available for teams that prefer their documented platform source and query language.

## Source documentation

- [Cisco ASA message definitions: 199001, 199002, and 199005](https://www.cisco.com/c/en/us/td/docs/security/asa/syslog/asa-syslog/syslog-messages-101001-to-199021.html)
- [Cisco FTD message definitions: 199001, 199002, and 199005](https://www.cisco.com/c/en/us/td/docs/security/firepower/Syslogs/fptd_syslog_guide/syslogs1.html)
- [Splunk Connect for Syslog: ASA/FTD source types and collection setup](https://splunk.github.io/splunk-connect-for-syslog/main/sources/vendor/Cisco/cisco_asa/)
- [Microsoft: Syslog via AMA ingestion](https://learn.microsoft.com/en-us/azure/sentinel/connect-cef-syslog-ama)
- [Microsoft: Syslog table fields](https://learn.microsoft.com/en-us/azure/azure-monitor/reference/tables/syslog)
- [Sumo Logic: matches operator and regex syntax](https://www.sumologic.com/help/docs/search/search-query-language/search-operators/matches/)
- [LogScale: query basics and field-specific regex](https://library.humio.com/crowdstrike-query-language/writing-queries-operations.html?redirected=true)
- [LogScale: native raw message field](https://library.humio.com/logscale-gdi/rawstring-field.html)

Source documentation inspected October 30, 2026. Review the current advisory before applying its release or configuration guidance.

- [Sigma specification: full-event keyword lists and string matching](https://sigmahq.io/sigma-specification/specification/sigma-rules-specification.html)
