containing-active-breach

Files (4)

• LICENSE
• SKILL.md
• api-reference.md · references/
• agent.py · scripts/

SKILL.md (first lines)

# Containing Active Breaches

## When to Use

- A confirmed intrusion is in progress with an active adversary on the network
- Malware is spreading laterally across endpoints or servers
- A compromised account is being used for unauthorized access to systems
- Ransomware encryption has been detected and is actively propagating
- An attacker has established command-and-control communications from internal hosts

**Do not use** for post-incident cleanup when the adversary is no longer active; use eradication procedures instead.

## Prerequisites

- Confirmed incident classification with P1 or P2 severity from triage
- EDR console access with host isolation capabilities (CrowdStrike Falcon, Microsoft Defender for Endpoint, SentinelOne)
- Network firewall and switch management access for segmentation
- Active Directory or identity provider administrative access for credential actions
- Pre-approved containment authority documented in the incident response plan
- Evidence preservation plan to avoid destroying forensic artifacts during containment

## Workflow

### Step 1: Assess Containment Scope

Before taking containment actions, map the full scope of compromise to avoid partial containment that alerts the adversary:

- Identify all confirmed compromised hosts via EDR telemetry and SIEM correlation
- Map lateral movement paths using authentication logs (Windows Event ID 4624 Type 3 and Type 10)
- Identify all compromised credentials (check for pass-the-hash, Kerberoasting, DCSync activity)
- Determine C2 channels (beacon intervals, domains, IPs, protocols)
- Assess whether the adversary has domain admin or equivalent privileges

```
Containment Scope Assessment:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Compromised Hosts:     5 (WKSTN-042, WKSTN-087, SRV-FILE01, SRV-DC02, WKSTN-103)
Compromised Accounts:  3 (jsmith, svc-backup, admin-tier0)
C2 Channels:           HTTPS beacon to 185.220.x.x every 60s ± 15% jitter
Lateral Movement:      PsExec via svc-backup, RDP via admin-tier0
Adversary Privilege:   Domain Admin (admin-tier0 compromised)
Data at Risk:          Finance share (\\SRV-FILE01\finance$) accessed
```

### Step 2: Execute Short-Term Containment

Implement immediate actions to stop adversary operations without destroying evidence:

**Network Containment:**
- Isolate confirmed compromised endpoints via EDR network containment (maintains agent communication)
- Block C2 IP addresses and domains at perimeter firewall and internal DNS
- Implement microsegmentation rules to prevent communication between compromised hosts
- Sinkhole C2 domains at internal DNS to capture connection attempts from undiscovered implants

**Identity Containment:**
- Disable compromised user accounts in Active Directory (do not delete; preserve audit trail)
- Reset passwords for all compromised accounts
- Revoke active sessions and tokens (Azure AD: `Revoke-AzureADUserAllRefreshToken`)
- Disable the compromised service account and rotate its credentials
- If Domain Admin is compromised: double-reset the KRBTGT password (reset twice, 12 hours apart)

**Endpoint Containment:**
- Use EDR to terminate malicious processes on contained hosts
- Block known malicious hashes in EDR prevention policy
- Quarantine identified malware samples
- Disable remote services (WinRM, RDP, SMB) on critical servers not yet compromised

### Step 3: Execute Long-Term Containment

Implement sustainable containment while the investigation continues:

- Create network ACLs isolating the compromised VLAN/subnet while allowing business-critical traffic
- Deploy temporary jump hosts for administrators to access contained systems for investigation
- Implement enhanced monitoring (full packet capture) on network segments adjacent to compromised hosts
- Enable advanced audit policies on all domain controllers (4768, 4769, 4771 for Kerberos attacks)
- Deploy canary tokens and honeypot accounts to detect adversary attempts to expand from containment

### Step 4: Validate Containment Effectiveness

Confirm that containment measures have stopped adversary operations:

- Monitor for new C2 callbacks from any internal host to known adversary infrastructure
- Check for new lateral movement attempts (failed authentication from disabled accounts)
- Verify that contained hosts cannot reach the internet except through the EDR agent
- Confirm that compromised credentials produce authentication failures
- Review SIEM for any new alerts matching the adversary's known TTPs

```
Containment Validation Checklist:
[x] C2 beacon traffic ceased from all known compromised hosts
[x] Disabled accounts producing expected 4625 failure events (no new successes)
[x] Contained hosts unreachable via network scan from adjacent subnets
[x] No new hosts exhibiting IOCs from the initial compromise
[x] Honeypot account has not been accessed (adversary may be dormant)
[ ] Full packet capture running on finance VLAN (pending switch config)
```

### Step 5: Preserve Evidence During Containment

Containment must not destroy forensic evidence:

- Capture memory dumps from compromised hosts before any remediation (use WinPmem or Magnet RAM Capture)
- Collect volatile data: running processes, network connections, logged-on users, scheduled tasks
- Export relevant event logs before they rotate (Security, System, PowerShell, Sysmon)
- Capture network traffic between compromised hosts and C2 infrastructure
- Document all containment actions with timestamps for the incident timeline

### Step 6: Communicate Containment Status

Provide structured status updates to incident commander and stakeholders:

- Current containment effectiveness (percentage of adversary activity stopped)
- Remaining risks (undiscovered implants, persistence mechanisms not yet identified)
- Business impact of containment actions (which systems are offline, user impact)
- Estimated timeline for eradication phase
- Escalation needs (law enforcement notification, external IR retainer activation)

## Key Concepts

From adriannoes/awesome-agentic-ai, Apache-2.0.

More from adriannoes/awesome-agentic-ai