OpenSecOps Newsletter

OpenSecOps Newsletter

Archives
Subscribe
November 8, 2026

OpenSecOps SOAR 4.3.0: Smarter Incidents from CloudWatch Alarms

SOAR 4.3.0 gives incidents from CloudWatch alarms their history and their cause. When an alarm on a Step Functions state machine or a Lambda function raises an incident, the AI analysis now knows how often that alarm has fired before and what exactly failed this time, so it can tell a one-off from a recurring problem and go straight to the error. This release also adapts OpenSecOps to an AWS change in cross-account event delivery, and adds two alarms so that a change like this is no longer invisible.

What's happened

AWS has changed how EventBridge authorises the delivery of events to an event bus in another account. Under the new rules, one of the OpenSecOps protection SCPs stopped Security Hub findings from being forwarded from the Security-Adm account to SOAR. We saw the change take effect on 11 August 2026 in eu-north-1. The date may differ in other Regions. Its effect was not easy to see: it showed only in the forwarding rule's metrics, which nothing watched. This release adapts the SCP to the new rules, so that findings are forwarded while the event bus and its rules stay protected.

What this release gives

Smarter incidents from CloudWatch alarms

The quickest way to create an incident in OpenSecOps SOAR is to name a CloudWatch alarm for it. SOAR watches every CloudWatch alarm in every account and Region it covers. Any alarm whose name contains a severity, INFORMATIONAL, LOW, MEDIUM, HIGH or CRITICAL, in capitals and set off by a hyphen, raises an incident of that severity when it fires. Orders-Queue-Backlog-HIGH is such a name. Include INFRA in the name, and the incident goes to the infrastructure team instead of the application team. There is no code to write and nothing to integrate: the name is enough. The convention is described in section 4.2 of the SOAR Technical Design Specification.

New in SOAR 4.3.0: when the alarm watches a Step Functions state machine or a Lambda function, the incident brings two kinds of context to the AI analysis.

  • History: how many times the same alarm has raised an incident in the last 7, 30, 90 and 365 days, when it first and last did, and whether this is a first occurrence, an isolated one, or part of an infrequent, moderate or frequent pattern.
  • The failure itself: for a state machine, the execution that failed, the state it failed in, the error and its cause, and the failing Lambda function with its error log lines. For a Lambda function, its error log lines from the time of the alarm.

With this, the analysis can tell the team whether it is looking at a one-off or a recurring problem, quote the actual error, and point its troubleshooting at the specific execution or function, instead of reasoning from the alarm name alone. The OpenSecOps infrastructure alarms on the SOAR state machines benefit too.

Alarm findings also record the exact time the alarm fired, and their description now includes CloudWatch's own reason for firing, such as the datapoint that crossed the threshold. Incident emails show that description above the AI analysis.

Alarms on the findings forwarding

Two new CRITICAL alarms watch the forwarding of findings to SOAR, so that a change like the one above no longer goes unnoticed. One fires on the first finding that cannot be delivered, the other when the forwarding has been idle for 24 hours. Both notify a new SNS topic, OpenSecOpsSOARFindingsForwardingFailures, which works independently of SOAR, so the alarms reach the administrators you subscribe to it even when SOAR receives nothing.

Also in this release

S3 tag-triggered incidents and log-bucket detection now work in every Region. One CVE is fixed: CVE-2026-49272 in oauthlib, in code that SOAR does not use.

How to install and configure it

  1. cd Installer && git pull, then ./init Foundation and ./init SOAR.
  2. Copy foundation/SCPs/protect-infra-immutable.json from apps.example/ to apps/. If you have customised your copy, replace the line "events:Put*", with "events:PutPermission",, "events:PutRule", and "events:PutTargets", instead.
  3. ./deploy-all Foundation, then ./deploy-all SOAR.
  4. Subscribe your AWS administrators to the SNS topic OpenSecOpsSOARFindingsForwardingFailures in the Security-Adm account, main Region. Email subscriptions become active only once the confirmation link has been clicked.
  5. The findings forwarding now runs in the main Region only, where Security Hub gathers the findings from all Regions. Earlier versions also deployed it to the other Regions, and those copies are superfluous. In the Security-Adm account, delete the CloudFormation stack INFRA-sec-hub-configuration in every Region except the main Region, using an administrator role that the protect-infra-immutable SCP exempts. The main-Region stack holds the forwarding rule, the two alarms and the topic. If it is deleted by mistake, the next deploy recreates it, but until then no findings reach SOAR and no alarm reports it, and the topic comes back without its subscriptions.

SOAR processes findings from other Regions only if Security Hub cross-Region aggregation into the main Region is enabled. The Foundation component Foundation-security-services-setup reports whether it is.

The released components are SOAR 4.3.0, SOAR-sec-hub-configuration 1.4.0 and SOAR-sec-hub-role 1.0.13, together with the already released Installer 3.1.1 and SOAR-all-alarms-to-sec-hub 1.3.8. Each component's CHANGELOG.md has the details.

Don't miss what's next. Subscribe to OpenSecOps Newsletter:
Older → OpenSecOps SOAR 4.0.0: Current Claude Models
GitHub
www.opensecops.org
LinkedIn
Powered by Buttondown, the easiest way to start and grow your newsletter.