Skip to content

Latest commit

 

History

History
61 lines (50 loc) · 3.09 KB

File metadata and controls

61 lines (50 loc) · 3.09 KB

Recommendations

These are the recommendations carried out of the engagement for a production version of this estate, ordered by horizon. Each one traces to something this engagement actually demonstrated.

Short term (under 30 days)

  1. Adopt a managed password vault. The single highest-impact day-one finding was one password reused across the entire estate. A shared vault removes the reuse pattern at the root.
  2. Enable multi-factor authentication on every administrative and domain account. None had it.
  3. Deploy LAPS (Local Administrator Password Solution) so local administrator passwords are randomized per host and centrally managed.
  4. Alert on telemetry agents going offline. In SO-2 the red team deleted the Security Onion agent to blind the SIEM, and the silence was misread as calm. An agent going dark must itself raise an alert.

Medium term (one to three months)

  1. Establish a patch cadence - monthly cumulative patching plus emergency patches within 72 hours of a relevant CVE. The estate carried roughly 17,600 vulnerability instances and multiple end-of-life operating systems on day one.
  2. Enforce DMZ-to-LAN segmentation with explicit allow rules. The flat topology meant a single web-application compromise had unrestricted reach into the internal network.
  3. Apply egress filtering. The reverse shell in SO-2 depended on an outbound callback to a command-and-control host; an egress allow-list would have broken the attack at that point.
  4. Remove world-writable upload directories and validate file uploads. The red team's initial foothold was a web shell dropped into a world-writable directory.

Long term (three to twelve months)

  1. Deploy behavioral EDR on every host, for detection beyond signature matching.
  2. Maintain an auto-discovered asset inventory. The deployed estate did not match the paperwork - addressing, segmentation, and host count all differed. An inventory sourced from discovery rather than manual entry keeps that gap from reopening.
  3. Run continuous vulnerability management. The configuration assessment and vulnerability scanner gave excellent visibility during the engagement; automated, tracked remediation is the missing half.
  4. Keep incident-response runbooks and rehearse them. Response under time pressure is a perishable skill. The unplanned domain-controller recovery on Day 3 went smoothly only because the change registry made it a rehearsed replay rather than an improvisation.

The through-line

The pattern across all twelve is the same: the engagement won where a control was in place before it was needed - the pre-cached blocklist, the rotated credentials, the detection layer, the change registry - and lost time wherever a control was reactive. Every recommendation above is a way of moving one more control from reactive to ready.


← Previous: Outcome and Forensics | Contents | Next: Appendix A - Change Log →