Workload Audit
and controlled response
Detect. Defend. Prevail.
A development checkpoint across the three Linux pilot hosts.
The three-host development fleet now includes scoped behavioral Audit, explicit process response, and more readable finding evidence. These are development milestones, not a public release.
Workload-aware behavioral Audit
The current policy assesses observed activity in the context of one configured, non-delegated system service per host. It checks service identity and observed child relationships, then correlates selected network, file, and execution observations into a reviewable finding.
Controlled checks use dedicated development services. The correlation remains post-execution Audit: it does not automatically terminate processes or block execution. Missing or ambiguous identity remains a visible gap, and network metadata does not establish transferred contents or remote file origin.
Operator-authorized process response
The pilot supports an explicit SIGTERM request for one verified live process. The operator inspects the current target, provides a reason, and confirms. Local CLI and console response remain available without a dashboard connection.
Receipts distinguish an acknowledged signal from a confirmed exit, a rejection, or an unknown outcome. Process response defaults off and is enabled for pilot testing. This slice does not include automatic response, process-tree termination, forced-kill escalation, network isolation, or quarantine.
Readable investigation evidence
Finding details now present readable filenames, paths, and event-family summaries beside the cited evidence. Retained identifiers and raw records remain available in technical details. Missing records and context limits remain explicit.
Current development work
Current work is focused on process-generation source migration, investigation refresh, and indexed-query efficiency. An isolated source-mechanism qualification is not a repaired installed collector; those migration and delivery gates remain open.
DarkAegis remains pre-alpha. Package lifecycle, representative workload precision, recovery, resource use, and public-release acceptance require further validation. Machine learning remains deferred from evaluator and alpha builds.
Discuss the project or a controlled evaluation.
DarkAegis
Detect. Defend. Prevail.