Chapter 03

Identification and Scoping

A single suspicious PowerShell event is not an incident. Understanding what you are actually dealing with — and how far it reaches — is the most consequential work in any IR engagement.

Scenario

An analyst investigates a single Sysmon alert: PowerShell making an outbound HTTPS connection from a finance workstation. Baseline enrichment shows the domain was registered 3 days ago. The analyst escalates. The IR lead begins scoping. Six hours later, the picture looks like this: the PowerShell was a Cobalt Strike beacon. The beacon had been running for 62 days. In those 62 days, the attacker had moved laterally to 200 hosts using a service account credential harvested on day 3. They had accessed 4 file servers containing financial data. They had two additional C2 channels that the original alert hadn't caught. They had created 3 backdoor accounts in Active Directory. None of this was visible from the original alert. All of it was visible once the analyst stopped triaging the single event and started scoping the full intrusion.

Security Event vs Confirmed Incident

Every IR begins with an event — a single data point that may or may not represent malicious activity. The decision to declare an incident is a threshold decision with real consequences: declaring too early wastes resources and builds alert fatigue; declaring too late allows damage to accumulate. Understanding the signals that separate a security event from a confirmed incident is one of the most important skills in IR.

  Event → Investigation → Incident Declaration Decision
  ══════════════════════════════════════════════════════════════════════

  SECURITY EVENT (a data point)
  ─────────────────────────────────────────────────────────────────
  Examples: IDS alert, suspicious process creation, failed logins,
  AV quarantine, phishing report, anomalous DNS query.

  Initial triage questions:
  • Is this alert a known false positive pattern for this environment?
  • Does enrichment (VirusTotal, passive DNS, threat intel) confirm malice?
  • Is there corroborating evidence in other data sources?

       └── NO corroboration, benign explanation likely → Monitor, close
       └── UNCERTAIN → Investigate further (not yet an incident)
       └── YES corroboration or no benign explanation → Escalate

  INVESTIGATION (gathering evidence to make the declaration decision)
  ─────────────────────────────────────────────────────────────────
  Goal: answer these four questions
  1. Is there confirmed malicious activity on this host?
  2. When did it start?
  3. What has it done (or had access to)?
  4. Is there evidence of spread to other hosts?

  Signals that force Incident Declaration:
  • Confirmed C2 beacon (network pattern + process evidence)
  • Confirmed credential theft (lsass.exe memory read, SAM export)
  • Confirmed lateral movement (authentication from compromised host
    to multiple others in unusual pattern)
  • Confirmed persistence (scheduled task, service, registry run key
    created by known-malicious process)
  • Confirmed data staging or exfiltration (large compressed archive
    created, large outbound transfer to unusual destination)

  INCIDENT DECLARED
  ─────────────────────────────────────────────────────────────────
  From this point: IR process is active. IR lead is engaged.
  Legal counsel is notified. Retainer is contacted.
  All actions are timestamped and documented.
Mental model: the event is the symptom, the intrusion is the disease

The alert that triggers an IR investigation is almost never the full story. It is the visible manifestation of attacker activity — the thing that broke a detection threshold. The actual intrusion may have been running for weeks or months before the first alert fired. The most important mindset shift in moving from security analyst to incident responder is: stop triaging the alert and start investigating the intrusion. The alert tells you where to look. The investigation tells you what is actually there.

Scoping: Mapping the Full Blast Radius

Scoping is the process of answering the question: how far does this intrusion reach? It is systematically expanding outward from the initial point of detection — patient zero — until you can confidently say that you have found every compromised system. This is harder than it sounds, because the attacker has been actively working to spread their access while you have been unaware of their presence.

  Scoping Methodology — Expanding Outward from Patient Zero
  ══════════════════════════════════════════════════════════════════════

  PATIENT ZERO (first confirmed compromised host)
       │
       ├── LOOK BACK IN TIME
       │   "When did this start?"
       │   • Check EDR telemetry for earliest malicious activity
       │   • Check authentication logs going back 90 days
       │   • Check DNS logs for C2 domain queries before the alert
       │   → Establishes the dwell time
       │
       ├── LOOK OUTWARD FOR LATERAL MOVEMENT
       │   "What did this host connect to?"
       │   • Authentication events FROM this host (Event 4624/4648)
       │   • SMB connections to other hosts
       │   • RDP sessions from this host
       │   • PowerShell remoting / WinRM from this host
       │   • WMI remote execution from this host
       │   → Every destination is a potential secondary compromised host
       │
       ├── LOOK FOR CREDENTIAL USAGE
       │   "What credentials were harvested from this host?"
       │   • Was lsass.exe accessed with PROCESS_VM_READ? (Event ID 10)
       │   • Was a credential dumping tool (comsvcs.dll, procdump) run?
       │   • What accounts were logged into this host? Those credentials
       │     are potentially compromised.
       │   → Every harvested credential defines a new blast radius
       │         (everywhere that credential was valid and used)
       │
       └── LOOK FOR C2 INFRASTRUCTURE
           "Are there other C2 channels I haven't found?"
           • DNS queries across the environment for the C2 domain
           • Other hosts with the same beacon process
           • Other hosts with the same parent process tree pattern
           → Identifies additional compromised hosts independently
             of the lateral movement chain

  For each newly identified compromised host:
  → Repeat the same expansion (look back, look outward, look for creds)
  → The blast radius grows until no new hosts are found

The Patient Zero Problem

Patient zero is the first host compromised in an intrusion. Finding it is essential for two reasons: it tells you the initial access vector (so you can close it during eradication), and it establishes the true start of the incident timeline. When an attacker has been in your environment for 60 days, patient zero may not be the host where you first detected activity — it may be a host the attacker compromised two months ago and used only as a staging point.

Initial access vectorHow to find patient zeroKey evidence
Phishing email with payload Email gateway logs: find the delivery of a suspicious email attachment or link in the timeframe before the earliest known compromise Email gateway logs, Process creation events showing document opening child process (Word → PowerShell)
Compromised credentials (password spray / credential stuffing) Authentication logs: find the first successful authentication from an unusual IP or location for a user account Event 4624 from unusual IP, Event 4648 explicit credential use, Azure/AAD sign-in logs with unusual location
Exploitation of internet-facing service (VPN, OWA, etc.) Web server / application logs for the exploited service: find unusual requests in the period before compromise HTTP access logs, WAF logs, IIS/Apache error logs, authentication events from the service
Supply chain (malicious update, compromised vendor) Identify the common software or service that all affected hosts share; correlate installation/update timestamps Software inventory, patch history, vendor access logs
Insider threat Activity originates from legitimate user context with no external compromise indicator; requires behavioral analysis DLP logs, unusual access patterns to sensitive data, after-hours activity, data staging and transfer

The Counterintuitive Advice: Scope Before You Contain

Under pressure, every instinct says: contain now. Stop the bleeding. Isolate the compromised host. This instinct is correct in limited cases — if there is active damage happening right now (files being encrypted, data being exfiltrated in real time), you stop it immediately. But in the more common case of a persistent intrusion with an attacker who has been present for days or weeks and is not actively causing damage at this moment, premature containment is a trap.

Why scope-before-contain leads to better outcomes

When you isolate the first compromised host before scoping, two things happen. First, you tip off the attacker that they've been found — sophisticated actors monitor their own beacons and immediately know when a host goes dark. Second, you trigger panic reactions: the attacker may detonate a destructive payload on the other 40 hosts they own that you haven't found yet, precisely because they know IR is coming and they want to destroy evidence and maximize damage before they're fully evicted. An attacker who has been quietly persistent for 60 days will not simply leave when you contain one host — they will use all the other access they have established, aggressively, before you can stop them. Spending 90 minutes on comprehensive scoping — mapping the full blast radius via log analysis before touching any host — gives you a simultaneous containment opportunity that denies the attacker the ability to respond. This is harder and requires more discipline than immediate containment, but it consistently produces better outcomes in complex intrusions.

  Sequential Containment vs Simultaneous Containment
  ══════════════════════════════════════════════════════════════════════

  SEQUENTIAL (contain as you find):
  ─────────────────────────────────────────────────────────────────
  T+0:00  Alert fires. Analyst confirms. Isolates host A.
  T+0:30  Attacker sees host A go dark. Pivots to host B.
  T+0:45  Analyst finds host B via lateral movement indicators.
          Isolates host B.
  T+1:00  Attacker pivots to host C (already compromised).
          Begins data staging.
  T+1:30  Analyst finds host C. But data has already been staged.
  T+2:00  Attacker uses host D to exfiltrate staged data before
          analyst gets there.
  Result:  3 hours, 4 containment actions, data exfiltrated.

  SIMULTANEOUS (scope first, contain all at once):
  ─────────────────────────────────────────────────────────────────
  T+0:00  Alert fires. Analyst confirms. Does NOT isolate yet.
  T+0:00  → T+1:30  Active scoping via log analysis and EDR queries.
          Find all 4 compromised hosts (A, B, C, D) plus 2 backdoor
          accounts and the C2 infrastructure pattern.
  T+1:30  Simultaneous isolation of A, B, C, D via EDR bulk contain.
          Simultaneous account disables.
          Simultaneous firewall block of C2 infrastructure.
  T+1:32  Attacker loses all access simultaneously. No pivot available.
  Result:  1.5 hours, 1 coordinated containment action, no exfiltration.

  Prerequisite for simultaneous: comprehensive visibility (EDR everywhere,
  centralized logs). The scoping investment only works if the data exists.

The Unknown Unknowns Problem

The hardest part of scoping is knowing when you're done. You can enumerate the compromised hosts you've found — but you cannot enumerate the hosts you haven't found because you don't know they exist. The only defense against this is systematic methodology: don't stop scoping because you ran out of obvious indicators, stop scoping because you have exhausted every lateral movement path from every confirmed host and found nothing new.

Scope closure signalWhat it meansWhat it doesn't mean
No new compromised hosts found in 48 hours of investigation The expansion loop from all confirmed hosts has produced no new findings That there are definitely no more hosts — only that you haven't found any more using current methods
Patient zero identified and initial access vector confirmed The intrusion timeline is closed at the beginning That you've found everything that happened between initial access and detection
All known attacker tools and techniques accounted for Every artifact associated with the identified actor's TTPs has been found or accounted for That the attacker didn't use techniques outside the known TTP set
Threat intelligence confirms scope of known campaign External intel corroborates your findings as consistent with the actor's typical scope That the actor didn't deviate from their typical pattern in this specific engagement
Common mistake: declaring scope closed because the initial hosts are clean

The most common scoping failure is stopping the expansion loop at the first ring of compromised hosts — cleaning those hosts and assuming the blast radius was limited to what was found. Attackers routinely establish footholds on systems well beyond the initial compromise before IR begins. A phishing victim's workstation is almost never the extent of the compromise — it is the entry point. The extent of the compromise is everywhere the attacker moved from that entry point, which in a mature intrusion is often dozens or hundreds of hosts. Scope closure requires exhausting the expansion methodology from every confirmed host, not just cleaning the hosts you found first.

When to Bring in External IR

External IR firms provide three things that internal teams frequently cannot: scale (they can deploy 10 analysts immediately where you have 2), specialized forensic capability (memory forensics, malware analysis, threat actor attribution), and independence (for legally sensitive incidents, an independent investigation is often required). The decision to engage is not about capability — it is about capacity and context.

IndicatorExternal IR likely needed
ScaleMore than 20 confirmed compromised hosts, or evidence the attacker has been present more than 30 days
ExpertiseNation-state actor suspected, or malware requiring advanced reverse engineering, or active memory forensics required
IndependenceIncident involves potential insider threat, executive accounts, or is likely to result in litigation or regulatory investigation
CapacityIR team is handling the incident and cannot simultaneously maintain normal security operations
InsuranceCyber insurance policy requires engagement of approved IR firm

Q & A

How do you determine how long the attacker has been in the environment when the earliest EDR telemetry is only 90 days old?

Log sources have different retention windows, and correlating across all of them extends your visibility backward. Even if EDR only goes back 90 days, authentication logs often go back longer. DNS passive logs may show the first query to a C2 domain. Email gateway logs may show the initial phishing email delivery. Firewall logs may show the first outbound connection to attacker infrastructure. If all logs are exhausted and you still cannot find the initial compromise, the answer is "we cannot determine the start date with certainty" — which is an honest and documentable finding. What you don't do is assume the attacker has only been present as long as your earliest log evidence shows. The attacker may have arrived before your logging began or before your current log retention window. "Earliest confirmed evidence is X date" is accurate; "attacker arrived on X date" may not be.

An executive is pushing to contain immediately — they want to "stop the bleeding." How do you explain why you're spending 90 minutes scoping first?

Frame it as risk management: "If we contain now, we isolate one host and tip off an attacker who owns 40 more. They will use those 40 hosts aggressively in the next 30 minutes before we can find and contain them. If we spend 90 minutes mapping every host they own and then contain all of them simultaneously, we eliminate their ability to respond. The 90-minute investment eliminates 3 days of reactive containment. The tradeoff: are the hosts they have now doing active damage right now, in the next 90 minutes? If yes, we contain immediately. If no — and our telemetry shows no active exfiltration or encryption — we scope first." Then commit to a timeline: "In 90 minutes I will have a complete picture and we will contain everything simultaneously." Executives respond to specific commitments, not to arguments about methodology. Give them a time and deliver it.

How do you track and communicate blast radius as it expands during active scoping?

A live document (not a mental model) updated in real time. Structure: confirmed compromised hosts (with the evidence that confirms them), suspected hosts under investigation (with the lead that triggered investigation), cleared hosts (investigated and confirmed clean), scope boundaries (network segments / user populations that have been investigated and show no evidence). Update this every 15–20 minutes during active scoping. Brief the IR lead after each update. Brief the CISO when the count crosses a threshold (10 hosts, 50 hosts, 100 hosts — pick thresholds in advance so you're not making that judgment call under pressure). The document serves two purposes: operational coordination (everyone knows what's confirmed vs suspected) and post-incident evidence (the timeline of how the blast radius was understood over time).

What is the difference between a host being "compromised" and a host being "exposed"?

A compromised host has confirmed evidence of attacker activity — malware installed, unauthorized authentication, persistence mechanism created, or credentials harvested. An exposed host is one that was reachable from a compromised host and had valid credentials used against it, but where active compromise has not yet been confirmed. Exposed hosts require investigation but should not be announced as compromised — that conflates certainty with possibility and may trigger premature or unnecessary incident response actions. In scoping documentation: compromised hosts go in the "confirmed" list, exposed hosts go in the "under investigation" list. When investigation of an exposed host finds no evidence of attacker activity, it moves to the "cleared" list. This distinction matters for legal communications: you may have mandatory notification obligations for confirmed compromised systems that process PII; you do not yet have those obligations for exposed-but-not-confirmed systems.

How do you scope a cloud intrusion differently from an on-premises intrusion?

The fundamental methodology is the same — expand from the initial indicator, follow credential usage and lateral movement — but the data sources and movement patterns differ. In cloud environments, lateral movement often happens through IAM: an attacker who compromises an EC2 instance may use its instance profile to assume other IAM roles, access S3 buckets, or create new IAM users. The analog of "what did this host connect to" becomes "what IAM actions did this principal take" and "what resources did it access." Key cloud lateral movement indicators: cross-account role assumptions from compromised accounts, new IAM user or access key creation, unusual API calls to high-privilege services (IAM, KMS, CloudTrail), and data access from EC2 instances to S3 buckets that don't normally communicate with them. Cloud logging retention is also different — CloudTrail defaults to 90 days if configured, but may not be configured at all in every account, and some data plane events (S3 object access) require explicit enablement of additional logging. Scoping cloud intrusions often reveals logging gaps that make comprehensive scope determination impossible, and that finding itself must be documented.

A host is showing indicators of compromise, but the business owner says it cannot be isolated because it supports a critical process. How do you handle this?

This is a risk decision, not a technical decision — it should be made by the appropriate business decision-maker with full information, not resolved by the IR team unilaterally. Present the risk clearly: "This host shows [specific indicators of compromise]. Leaving it connected means the attacker retains access to it, can use it as a pivot point to other systems, and may be able to access or exfiltrate [data it contains]. The risk of not containing is [specific]. What business process depends on it, and what is the impact of a 4-hour isolation window?" Then find alternatives: can the critical process be moved to a clean host during the isolation? Can a forensic copy be taken and the host rebuilt on a clean image in under 4 hours? Can network-level containment (firewall rules blocking lateral movement from this host while keeping the business-critical service available externally) reduce risk without full isolation? Document the decision and who made it. If the host is later used as a pivot to compromise additional systems, the documentation protects the IR team and shows the decision was made with full information by an authorized decision-maker.