Active Adversary Playbook
When you know an attacker is in your environment right now and is actively operating, the response calculus changes completely. You are no longer reconstructing history — you are in a live counterintelligence operation where every action you take may be observed, and every action you delay may cost more access.
A security analyst places a canary file — a fake document labeled "Q3_Acquisition_Targets.xlsx" — in a finance directory as part of a deception framework deployment. The canary file generates an alert whenever it is opened. Twenty minutes after placement, the alert fires.
The IR team now knows with high confidence: an attacker has access to the finance file server, is actively exploring file shares right now, and is interested in M&A data. They do not yet know how the attacker got in, how long they've been present, what else they've accessed, or where their persistence lives.
The question on the table: act now and potentially lose intelligence, or observe and learn more but risk the attacker taking destructive or exfiltration action?
The Psychological Dynamic: You Are Watching Someone Who May Know They're Watched
Active adversary incidents have a fundamental quality that separates them from incident response against historical evidence: the adversary is operating in real time. Sophisticated threat actors monitor their own operational security — they watch for signs that defenders are aware of them. Actions you take in response to an active adversary may be visible to that adversary.
What the Active Adversary Can Observe
═══════════════════════════════════════════════════════════════════
IF the attacker has deep access:
─────────────────────────────────────────────────────────────────
→ They may be reading your email (including incident communications)
→ They may be watching security tool alerts (if they have admin access
to your SIEM or security tools)
→ They may be monitoring Active Directory for accounts being added
to incident response groups
→ They will notice: unusual admin logins at odd hours, mass password
resets, host isolation events, sudden disabling of user accounts
IF the attacker has moderate access:
─────────────────────────────────────────────────────────────────
→ They likely cannot see your internal communications or security tools
→ They WILL notice: their current session terminating, their C2 going
dark (firewall block), account password changes for compromised accounts,
hosts disappearing from the network (isolation)
IF the attacker has limited access:
─────────────────────────────────────────────────────────────────
→ Limited visibility into your response
→ Will notice: their specific access path being closed
→ May have backup persistence they haven't used yet
IMPLICATION: At the start of active adversary response, assume the
worst case — that the attacker can see your incident communications.
This drives the requirement for out-of-band communication immediately.
The Core Decision: Go Loud or Stay Quiet
Every active adversary incident requires an early decision about response mode. There is no objectively correct answer — the right choice depends on specific factors that must be assessed quickly. Both modes have significant costs.
| Go loud (immediate full containment) | Stay quiet (observe and collect) | |
|---|---|---|
| What you do | Immediately isolate affected hosts, reset credentials, block C2, evict the attacker. Do it all at once. | Monitor attacker activity closely. Collect evidence of their methods, persistence, and targets. Contain only when you have full picture or must act. |
| Advantages | Stops ongoing damage and exfiltration immediately. Simpler to execute. Clear and decisive. | Learn the full attack chain: all persistence mechanisms, all compromised credentials, full scope of attacker access before they know you're watching. |
| Disadvantages | You may miss persistence mechanisms you haven't found yet. Attacker may use backup persistence to re-enter. If you only find 80% of the compromise and cut off the visible 80%, the remaining 20% lets them rebuild. | Attacker continues operating during observation. They may exfiltrate more data, escalate privileges, install additional persistence, or damage systems. Legal and regulatory risk: observation during known data exfiltration may be seen as failure to mitigate. |
| When to choose | Active data exfiltration in progress. Attacker appears close to end goal (ransomware staging, destructive action). You have legal/regulatory obligation to act. Business impact of continued access is high and immediate. | Attacker is in early reconnaissance phase. No data exfiltration in progress. You want to identify all persistence before eviction. You have time and legal space. You want to develop threat intelligence for attribution or future defense. |
The most common reason for re-compromise after eviction is evicting based on what you found, not based on what the attacker installed. A sophisticated attacker establishes multiple persistence mechanisms — their primary C2 implant, a backup DLL hijack on a different host, an OAuth app grant that survives password resets, a scheduled task on a domain controller with a different name. During active adversary response, if you contain based on the visible C2 without finding the backup mechanisms, the attacker simply switches to the backup and continues from where they left off. The value of the stay-quiet phase is watching the attacker use their full toolkit, which reveals their backup persistence before you need to evict.
Deception Techniques During Active Adversary Response
Deception is the controlled provision of false information to an active adversary. During active adversary response, deception serves two purposes: buying time for the IR team to complete scope assessment and eradication preparation, and maneuvering the attacker into a position where their actions are more observable and less harmful.
Deception Techniques and Their Applications ═══════════════════════════════════════════════════════════════════ CANARY FILES AND DIRECTORIES ───────────────────────────────────────────────────────────────── Deploy: fake documents with names that match attacker objectives (acquisition targets, password lists, financial records, executive emails) Alert: any access or download fires immediate alert Purpose: detect attacker presence, identify which data they're targeting, understand attacker's current location in the file system Example: The canary in this chapter's scenario alerted within 20 minutes of placement — giving the IR team precise attacker location CANARY CREDENTIALS ───────────────────────────────────────────────────────────────── Deploy: fake administrator accounts with realistic names and histories (never give them real access) — these accounts exist only in Active Directory and credential stores. No real user uses them. Alert: any authentication attempt with canary credentials = attacker has credential dumps and is trying stolen creds Purpose: detect credential harvesting and cracking activity HONEYPOT SYSTEMS ───────────────────────────────────────────────────────────────── Deploy: fake server appearing to hold high-value data (fake "finance-reporting-server" with fake reports) Alert: any connection or authentication to honeypot systems Purpose: detect lateral movement; understand attacker navigation EVIDENCE PRESERVATION DECEPTION ───────────────────────────────────────────────────────────────── When you intend to go loud shortly: before acting, deploy additional monitoring to the specific host where attacker is active. Capture a memory image if possible. Then act. Purpose: capture maximum forensic evidence before eviction DELIBERATE DELAYS ───────────────────────────────────────────────────────────────── If attacker is attempting to exfiltrate via a specific channel: introduce network slowdowns on that path (with leadership approval) Purpose: slow exfiltration while you prepare to cut access; reduce data volume exfiltrated Warning: requires precise execution; a mistake can alert attacker
Limits on Observation: Legal and Ethical Constraints
The decision to observe an active attacker rather than immediately contain them has legal dimensions. In some jurisdictions, knowingly allowing an attacker to continue accessing systems — especially systems containing personal data — can create regulatory liability. Legal counsel must be involved before choosing the stay-quiet mode for any significant duration.
| Constraint | Implication for stay-quiet |
|---|---|
| GDPR (EU) — breach notification within 72 hours of awareness | Once you have awareness of a breach likely affecting EU personal data, the 72-hour clock starts. Observation mode does not pause this clock. Notify first, observe after — or contain within the notification window. |
| HIPAA (US healthcare) — "without unreasonable delay" | PHI breaches require notification. Choosing to observe for days after discovering an active adversary with access to PHI is likely "unreasonable delay." |
| Wiretap / Electronic Surveillance Laws | Monitoring attacker communications on your own network is generally legal (you own the network). Monitoring attacker communications on third-party infrastructure requires legal authority. Don't try to hack back to observe the attacker on their C2. |
| Duty of care to employees and customers | If the attacker is accessing personal data of employees or customers, prolonged observation during that access may create civil liability. Legal counsel makes this judgment call. |
Threat Intelligence Role in Active Adversary Response
Threat intelligence transforms raw indicators (C2 IP, malware hash) into strategic context (who is this actor, what do they want, what have they done to other organizations). During active adversary response, that context directly affects decisions.
Threat Intelligence Questions During Active Adversary Response
═══════════════════════════════════════════════════════════════════
Identity questions (for decision-making):
─────────────────────────────────────────────────────────────────
Q: What threat actor does this malware/C2 infrastructure suggest?
→ Affects: stay-quiet vs. go-loud decision
Ransomware affiliate: time to deployment may be days, so
observation window is short
Espionage actor: may have been present for months; they are
patient; there is more time to observe
Q: What is this actor's end goal?
→ Ransomware operator: the end goal is deployment. Go loud before they deploy.
→ Espionage actor: the end goal is persistent access and data collection.
They will not deploy ransomware; observation is more viable.
→ Hacktivist: end goal may be destructive or data leak. Time-sensitive.
Q: Has this actor targeted similar organizations?
→ Their playbook in those organizations predicts their playbook here.
If they used GPO-based ransomware deployment in 3 prior victims,
assume they'll do it here too.
Q: What data did they exfiltrate in prior victims?
→ Identifies which data in your environment is at highest risk and
should be most closely monitored during observation.
Operational questions (for detection):
─────────────────────────────────────────────────────────────────
Q: What IOCs are associated with this actor?
→ Deploy IOC-based detection immediately
→ Monitor for all known C2 infrastructure, not just the single
C2 beacon you've identified
Q: What TTPs does this actor use for persistence?
→ Front-load your persistence hunt with their known techniques.
If they are known for Kerberoasting + scheduled task persistence
+ WMI subscriptions, check all three immediately.
Eviction Preparation: Do It Once, Do It Right
When the decision is made to go loud, preparation determines success. A well-prepared eviction takes all actions simultaneously to prevent the attacker from pivoting to backup persistence or alerting their operators. A poorly coordinated eviction that happens in stages gives the attacker time to respond between steps.
Eviction Preparation Checklist
═══════════════════════════════════════════════════════════════════
Before you act — have everything staged and ready:
─────────────────────────────────────────────────────────────────
[ ] All compromised hosts identified and listed (not some — all)
[ ] All persistence mechanisms documented (scheduled tasks, registry
keys, DLL hijacks, WMI subscriptions, OAuth app grants, etc.)
[ ] All compromised credentials identified and password resets staged
in the identity system (ready to execute simultaneously)
[ ] C2 block rules staged in firewall/proxy (not yet applied)
[ ] EDR bulk contain list prepared (all in-scope hosts)
[ ] Team assignments: who executes which action
[ ] Communication channel confirmed (out-of-band)
[ ] Start time coordinated (execute simultaneously, not sequentially)
[ ] Evidence capture: memory images from key hosts captured before eviction
[ ] Leadership briefed on "go" time
Execution sequence (simultaneous):
─────────────────────────────────────────────────────────────────
Second 0: Block C2 at firewall/proxy (cuts attacker's feedback)
Second 0: EDR bulk contain all in-scope hosts (cuts lateral movement)
Second 0: Mass credential reset for all compromised accounts (cuts auth)
Second 30: Verify containment (check EDR console for isolated hosts)
Minute 2: Remove persistence mechanisms on each host
Minute 10: Confirm attacker has no path back into the environment
Minute 30: Begin scope validation and eradication confirmation
Q & A
Q: The canary fired but we have no idea how the attacker got into the finance server. Should we act before knowing the entry point?
This is one of the most common active adversary dilemmas. The answer depends on what the attacker appears to be doing. If they are actively exfiltrating data: act now and scope the entry point afterward — stopping the exfiltration is more important than understanding how they got there. If they are in a reconnaissance phase (reading files, running enumeration tools, but not exfiltrating): you have some time. Spend 30-60 minutes in parallel: (1) Contain the finance server from the rest of the network to limit lateral movement while keeping it accessible for monitoring. (2) Aggressively scope the entry point using authentication logs and EDR telemetry. (3) Look for C2 connections in network telemetry from the finance server or hosts the attacker was on before reaching the finance server. You want to act with enough knowledge to make the eviction stick. Acting against only the visible host while the attacker has undetected persistence elsewhere produces re-compromise within hours.
Q: We chose stay-quiet for 8 hours to scope the full intrusion. The attacker just started an rclone exfiltration. Do you change modes?
Yes — immediately. Active data exfiltration is the hard boundary for stay-quiet observation. The cost-benefit changes when real data is leaving the environment in real time: every additional minute of observation is additional data exfiltrated. Activate the pre-staged eviction plan immediately. If the eviction plan is not yet fully staged (persistence not fully scoped), accept an incomplete eviction rather than allow continued exfiltration — you can deal with re-compromise; you cannot un-exfiltrate data. When making this call to leadership, frame it clearly: "We were in observation mode to maximize scope, but we've seen active exfiltration begin. We're acting now. We may have to deal with reinfection from unidentified persistence, but stopping the data loss takes priority."
Q: Leadership wants to "hack back" — access the attacker's C2 infrastructure to see what data they stole. Is this legal?
In the United States, almost certainly not without explicit legal authority. The Computer Fraud and Abuse Act (CFAA) prohibits unauthorized access to computer systems — including systems owned by attackers. "They attacked us first" is not a legal defense. There is an active law and policy debate about "active defense" and hackback legislation, but as of 2026, unauthorized access to third-party systems — even an attacker's C2 infrastructure — exposes the organization (and potentially individuals) to CFAA criminal liability. The practical path for accessing attacker infrastructure is through law enforcement cooperation: share your evidence with the FBI, which has legal authority to investigate and access attacker infrastructure. If this is a nation-state attack, CISA and relevant government agencies have authorities that your organization does not.
Q: During observation, the attacker logs in from an IP that resolves to a known Tor exit node. Does this tell us anything useful?
It tells you this attacker values operational security and routes their traffic through anonymizing infrastructure — which is consistent with sophisticated financially-motivated or nation-state actors. It does not tell you who the actor is or where they are physically. Tor exit nodes are used by both legitimate privacy-conscious users and threat actors. For your response, Tor exit node attribution means: (1) Source IP blocking will be ineffective — they'll use a different exit node. (2) Geographic attribution from IP is unreliable. (3) The actor is sophisticated enough to care about attribution evasion, which suggests they're not an unsophisticated commodity attacker. For threat intel purposes, correlation of the specific Tor circuits used with other incidents can sometimes provide attribution signals — share IOCs with threat intel community resources (ISAC, trusted partner organizations) to check for overlapping infrastructure.
Q: A threat intelligence report says this actor typically has 3-5 persistence mechanisms. We've found two. Can we go loud?
The missing persistence mechanisms are the critical risk here. If you evict the attacker before finding all 5, the undiscovered mechanisms become the re-entry path. The observation window should continue until you have high confidence that you've found all persistence, or until the cost of continued observation exceeds the benefit (active exfiltration, destructive action risk, legal obligation). Practical approach: use the threat intel report's known persistence techniques as a checklist. Actively search for each class of persistence (scheduled tasks, WMI subscriptions, registry Run keys, DLL hijacks, service creation, OAuth apps, domain controller-specific persistence) on all in-scope hosts. You may find the remaining 3 through proactive hunting rather than waiting for the attacker to use them. Finding persistence through hunting is preferable to waiting for the attacker to activate it during or after eviction.