IR Frameworks
47 days. 14,000 alerts. Nobody caught the attacker. PICERL, NIST 800-61, and SANS give you a vocabulary for incident response — but the moment you confuse the vocabulary for the work, you lose.
47 days. That's how long the attacker spent inside the network before anyone noticed. During those 47 days, the SIEM fired 14,000 alerts. The SOC triaged every one of them. None of the 14,000 caught the attacker — because every alert was IOC-based, and the attacker used infrastructure that had never been seen before. The breach was eventually discovered because a single analyst, working an unrelated ticket, noticed a domain controller making precisely timed outbound HTTPS connections to a domain registered three days before the attack.
After the breach was confirmed, the IR lead called the team together and said: "We're in the identification phase. We need to move to containment." An hour later, they were debating whether certain actions were containment or eradication. The frameworks they had practiced in tabletop exercises were consuming mental bandwidth that should have been spent on a single question: what can the attacker still do right now, and what are we doing to stop it?
Why Frameworks Exist — and Where They Break Down
IR frameworks exist because incident response without structure collapses under pressure. At 2 AM with an active attacker in your environment, decision fatigue is real. The IR lead cannot simultaneously think about forensic evidence collection, stakeholder communication, legal notification timelines, containment sequencing, and attacker TTPs. Frameworks distribute that cognitive load by providing a pre-agreed vocabulary and a set of questions to ask at each stage.
The breakdown comes when teams treat the framework as the goal rather than the work it represents. "Phase-tracking" — spending energy classifying every action into the correct PICERL bucket — is a failure mode that looks like rigor. The framework is a checklist. The attacker is the enemy. Keep those priorities in order.
The most valuable use of PICERL is before any incident happens: use it to audit your preparation (have you done all the Preparation-phase work?), design your runbooks (what does your Identification process actually look like?), and structure your post-mortems (what Lessons Learned actions are actually changing your next response?). During an incident, you manage by tracks and priorities — not by phase names. The practitioner who spends an incident asking "what phase are we in?" is optimizing for the wrong variable.
PICERL in Operational Detail
PICERL is the SANS framework: Preparation, Identification, Containment, Eradication, Recovery, Lessons Learned. It is the most widely cited model in the industry and the one you will see on every GIAC exam and every IR job description. Understanding it deeply means understanding not just what each phase is, but what the operational translation looks like — the exact questions you ask, the evidence you look for, and the judgment calls each phase demands.
PICERL — Operational Translation
══════════════════════════════════════════════════════════════════════
P — PREPARATION
─────────────────────────────────────────────────────────────────────
What you do: Write runbooks. Deploy tooling. Establish contacts.
Run tabletop exercises. Maintain asset inventory.
Key question: "When the incident happens, are we ready to respond?"
Done BEFORE the incident. If you're building this during the incident,
you have already failed Preparation.
I — IDENTIFICATION
─────────────────────────────────────────────────────────────────────
What you do: Confirm it's a real incident. Scope it. Determine severity.
Key questions: "Is this a true positive?"
"What systems are confirmed compromised?"
"Since when has the attacker been in?"
"What is the blast radius?"
Ends when: You have a confirmed incident with a scoped blast radius.
Common failure: Moving to containment before scoping is complete.
C — CONTAINMENT
─────────────────────────────────────────────────────────────────────
What you do: Stop the bleeding. Prevent further compromise.
Short-term: Isolate confirmed compromised hosts. Disable compromised
accounts. Block known C2 infrastructure.
Long-term: Segmentation, credential rotation, patch deployment.
Key question: "Can the attacker still move laterally right now?"
Critical rule: Containment must not destroy evidence you haven't
collected yet.
E — ERADICATION
─────────────────────────────────────────────────────────────────────
What you do: Remove ALL attacker presence. Every persistence mechanism.
Every backdoor account. Every piece of malware.
Key question: "Is there any path the attacker could use to regain access
without re-compromising from the outside?"
Common failure: Finding one persistence mechanism and stopping there.
Attackers plant secondary persistence expecting IR to
find and remove the obvious one.
R — RECOVERY
─────────────────────────────────────────────────────────────────────
What you do: Restore systems to normal operations from verified clean state.
Key question: "Are we confident these systems are clean before we reconnect?"
Critical rule: Heightened monitoring during recovery window. Attackers
frequently return to check if they've been evicted — and
that return attempt is a detection opportunity.
L — LESSONS LEARNED
─────────────────────────────────────────────────────────────────────
What you do: Post-mortem. Root cause. Detection gaps. Hardening.
Key question: "What specific changes prevent or detect this faster
next time?"
Output: Updated runbooks. New detection rules. Hardening tasks
with owners and deadlines.
NIST SP 800-61 — The Program-Management View
NIST Special Publication 800-61 (Computer Security Incident Handling Guide) is the US federal standard and is referenced in most compliance frameworks. It uses four phases instead of six, but more importantly it models the process as a cycle rather than a linear progression — Lessons Learned explicitly feed back into Preparation for the next incident.
NIST 800-61 Incident Response Lifecycle
══════════════════════════════════════════════════════════════════════
┌─────────────────────────────────────────────────────────────────┐
│ │
│ Preparation ◄──────────────────── Post-Incident Activity │
│ │ ▲ │
│ ▼ │ │
│ Detection & (Lessons feed │
│ Analysis ─────────────────────────────► back into │
│ │ preparation) │
│ ▼ │
│ Containment, Eradication & Recovery │
│ │
└─────────────────────────────────────────────────────────────────┘
NIST phases map to PICERL:
───────────────────────────────────────────────────────────────────
Preparation → P (Preparation)
Detection & Analysis → I (Identification)
Containment/Erad/Recovery→ C + E + R (bundled — tightly coupled work)
Post-Incident Activity → L (Lessons Learned)
Key additions NIST provides that PICERL doesn't:
───────────────────────────────────────────────────────────────────
• Incident prioritization criteria: functional impact (what is still
working?), information impact (what data was accessed?),
recoverability (can we restore, and how fast?)
• Documentation requirements: chain of custody, evidence inventory,
decision log
• Legal/regulatory obligations: notification timelines, coordination
with CISA, FBI, sector-specific ISACs
• External coordination: IR retainer engagement, law enforcement
liaison, threat intelligence sharing
NIST bundles these three activities because in practice they are tightly coupled and frequently run concurrently. On the first contained host, you may begin eradication (removing persistence) while still performing containment on twenty other hosts in the same incident. Recovery of lower-priority systems may begin while eradication is still ongoing for higher-priority ones. PICERL's separation into three distinct phases is useful as a checklist — it ensures you ask the right questions for each type of work — but it can create a false impression that they happen sequentially. They don't.
Reality: IR is Parallel, Not Sequential
Both PICERL and NIST present IR as a sequence of phases. Every real large-scale incident violates this model. During an active intrusion, the following tracks run simultaneously from the moment the incident is declared:
Parallel IR Tracks (real large-scale incident)
══════════════════════════════════════════════════════════════════════
FORENSICS TRACK CONTAINMENT TRACK COMMUNICATIONS TRACK
──────────────── ───────────────── ────────────────────
Collect memory Scope confirmed Brief CISO (live)
from patient zero hosts via EDR
│ │ Draft SITREP #1
Analyze initial Isolate Tier 0 for leadership
compromise vector systems first
│ │ Notify legal counsel
Find lateral Hold suspected re: notification
movement path hosts in monitoring obligations
│ │
Determine Coordinate with Out-of-band comms
blast radius IT on service established
│ dependencies (Signal / OOB channel)
Identify all │
persistence Rotate domain SITREP #2 (T+2h)
mechanisms admin creds
│
Network containment
at firewall/DNS
FORENSICS TRACK LEGAL TRACK THREAT INTEL TRACK
──────────────── ──────────── ──────────────────
Collect from Notification Identify malware
all confirmed hosts deadline tracking family / actor
│ │ │
Build attack Regulatory Pull threat intel
timeline reporting prep on actor TTPs
│ │ │
Report blast Insurance Identify other
radius update notification likely targets
in environment
The IR lead coordinates all tracks. Phases describe types of work,
not time ordering. The Lessons Learned phase begins in parallel
(log everything) even while active containment is ongoing.
The Real Question Is Not "What Phase Are We In?"
In the fog of a real incident, the framework provides scaffolding but it can also become a distraction. Teams that are phase-tracking at the expense of adversary-tracking are optimizing for the wrong thing. The question that should dominate every IR decision is:
"What can the attacker still do right now, and what are we doing about it?"
Every action taken during an incident should either answer that question or reduce the answer's scope. A containment action answers it by reducing what the attacker can still do. A forensic finding answers it by expanding your knowledge of their current capabilities. A communication action answers it by ensuring the right people have the information they need to make decisions. Phase classification serves none of these goals.
A war room full of people generating tickets and running containment actions looks like progress. But if each action is reactive — responding to what the attacker does rather than systematically mapping and cutting their access — the IR team is always one step behind. The attacker acts, the team reacts. The attacker pivots, the team re-scopes. This pattern can continue for days. The alternative: treat the first two hours as a scoping investment. Resist the pressure to "do something" until you understand what you're actually dealing with. An IR team that spends ninety minutes mapping the full blast radius before touching anything will outperform a team that starts containing immediately and keeps finding new compromised hosts for a week.
Checkbox IR vs Thinking IR
Checkbox IR means executing framework steps without understanding the specific attacker and incident in front of you. It is the most dangerous failure mode because it looks exactly like competence. The checklists get completed, the phases get marked, the ticket gets closed — while the attacker is still in the environment.
| Checkbox IR behavior | What it misses | Thinking IR alternative |
|---|---|---|
| Found one persistence mechanism → moved to eradication | Attackers plant secondary and tertiary persistence specifically anticipating IR. Finding the first is not finding all. | Enumerate all persistence locations systematically (Autoruns, scheduled tasks, services, WMI, registry run keys, startup folders, BITS jobs) before declaring eradication complete. |
| Reimaged the compromised host → closed the ticket | If you don't understand how they got in, they'll use the same path again. Eradication without root cause is cleaning the symptom, not the disease. | Root cause analysis before recovery. The compromise path must be understood and closed, not just the infected endpoint. |
| Reset the compromised user's password → contained | The attacker moved laterally before you found the account. Other accounts they accessed are still valid. Accounts they created are still active. | Map all authentication events from the compromised account for the last 90 days. Every system touched is a candidate for secondary compromise. |
| Blocked the C2 IP in the firewall → C2 severed | C2 infrastructure rotates. One blocked IP is a speed bump. Without confirming no active session exists at the host level, the attacker may still have an active callback. | Confirm via EDR telemetry that no active C2 process or network connection exists on any affected host. Block at the firewall is a supplement, not a substitute. |
| Recovery complete, ticket closed | Attackers return to check if they've been evicted. The recovery window is an active threat period, not a wind-down period. | Heightened monitoring for 30 days post-recovery. Any authentication from previously compromised accounts gets immediate review. |
In a documented APT intrusion, an IR firm was engaged three separate times over eight months by the same organization. Each engagement found and removed the attacker's presence. Each time, the attacker returned within weeks. On the third engagement, the IR firm's forensic team found a WMI subscription that had survived all three previous remediations — not because the previous teams missed it, but because the attacker was re-establishing it remotely within hours of each eradication, using a service account credential that was never rotated because it controlled a critical legacy application. The lesson: eradication and credential rotation are inseparable. Removing persistence without rotating every credential the attacker touched is not eradication — it is a temporary inconvenience.
How to Use a Framework Correctly
Frameworks are most valuable at three specific moments: before the incident (preparation audit), during the post-mortem (structured lessons-learned), and when training new analysts (giving them a mental map of the work). They are least valuable during the active incident itself, where they should function as background checklists rather than operational guides.
| When | How to use the framework | What to avoid |
|---|---|---|
| Before any incident | Preparation audit: Have you completed every item in the Preparation phase? Runbooks written? Tools deployed? Contacts established? Tabletops run? | Treating it as done once. Preparation degrades — runbooks go stale, tools lose coverage, contacts change. |
| During declaration | Use the Identification phase questions: Is this a true positive? What is the confirmed scope? What is the severity level? Who needs to be notified? | Using phase names to communicate status. "We're in containment" tells leadership nothing useful. "We have 12 confirmed compromised hosts, C2 is severed on 8, 4 are still being investigated" tells them everything. |
| During active IR | Track by work type, not phase: forensics track, containment track, comms track, legal track. Each track has an owner and a current status. | Phase debates. "Is this containment or eradication?" is a waste of everyone's time. Do the work. Label it later if the report requires a label. |
| After the incident | Lessons Learned structured by framework: What failed in Preparation? What gaps in Identification delayed discovery? What containment gaps allowed further spread? | Blameless post-mortems that produce no action items. Every lessons-learned session must produce specific, assigned, deadlined improvements. |
The most common failure in post-incident reviews is producing observations without owners or deadlines. "We need better logging" is an observation. "By Q3, the security engineering team will deploy Sysmon to all 1,200 endpoints using the SwiftOnSecurity baseline, with compliance verified monthly" is an action item. The difference between organizations that improve after incidents and organizations that repeat the same incidents is not the quality of their post-mortem discussions — it is whether those discussions produce work items that get tracked, assigned, and completed.
Q & A
NIST 800-61 bundles Containment, Eradication, and Recovery into one phase. PICERL separates them. In practice, which model is right?
Neither is "right" — they describe the same work from different perspectives. NIST bundles them because in large incidents they genuinely overlap: you begin eradication on the first contained host while still containing others, and recovery of non-critical systems may begin while eradication is still ongoing elsewhere. PICERL separates them to give practitioners distinct checklists — each phase has different success criteria and different questions to ask. The practical approach: use PICERL's separated phases as checklists for completeness (did we actually finish containment before starting eradication? did we actually verify clean state before recovery?), while managing the work as concurrent tracks rather than a strict sequence. The phase checkboxes prevent shortcuts; the parallel track model reflects how the work actually flows.
How do you determine when Identification ends and Containment should begin?
They overlap, and the right answer depends on what the attacker is doing right now. If there is active damage occurring — active encryption (ransomware), active data exfiltration, active lateral movement with new hosts being compromised — you begin short-term containment of the confirmed-active systems immediately, even while scoping continues. If there is no active damage — the attacker has established persistence and is waiting — you invest more time in scoping before touching anything, because premature containment will tip the attacker into destructive action before you understand the full blast radius. The practical trigger is: "Can I contain this now without (a) destroying evidence I haven't collected, and (b) triggering a destructive payload the attacker has pre-positioned?" If yes to both, contain. If uncertain, scope aggressively for the next 60 minutes before acting.
A manager declares the incident in the Recovery phase and wants to close the ticket. You believe eradication is incomplete. What do you do?
Push back with specifics, in writing. "We are not ready to close because we have not confirmed removal of persistence on hosts [list]. Specifically, [evidence] suggests WMI subscriptions may still be active on [specific hosts]. Premature closure means the attacker retains access and the incident will reopen — with additional damage and a higher remediation cost." Provide a specific action and timeline: "I need 48 hours to complete the persistence sweep. Then we can close." If leadership overrides you and closes the incident, document your objection in the incident record with your specific technical reasoning. When the attacker returns — and they often do — the documentation protects you and provides the next IR team with an accurate starting point. Never verbally object and let it go. The written record matters.
What is the difference between functional impact, information impact, and recoverability in NIST's incident prioritization model?
These are the three dimensions NIST uses to assign priority levels to incidents, and understanding them helps explain why identical technical events can have different response urgency. Functional impact measures how much the organization's ability to operate is currently degraded — a ransomware attack that encrypted payroll systems has higher functional impact than one that encrypted an employee's personal documents. Information impact measures what data may have been accessed or exfiltrated — an attacker who read customer PII has higher information impact than one who accessed internal project documentation. Recoverability measures how quickly and easily the organization can return to normal — a well-backed-up environment hit by ransomware has higher recoverability than one with no offline backups. These three dimensions combine to create an overall severity. An incident with high functional impact but high recoverability (everything is backed up) may actually be lower priority than an incident with lower functional impact but catastrophic information impact (a quiet data exfiltration of all customer credentials).
The PICERL model implies IR is linear. What actually happens in a large-scale APT intrusion?
In a large-scale intrusion, the Identification phase never truly ends until the incident is fully closed — new compromised hosts keep being discovered throughout containment and eradication. Containment begins before Identification is complete (you start isolating confirmed hosts while scoping continues). Eradication and Recovery run concurrently (recovery of confirmed-clean systems while eradication continues on others). And Lessons Learned begins in parallel with active IR — you're building the post-mortem timeline in real time even while containment is happening. The accurate model is: IR is a project with multiple concurrent workstreams, each at a different stage simultaneously, overseen by an IR lead whose job is coordination rather than execution. The linear framework is a useful fiction that helps new practitioners understand the types of work involved; experienced practitioners know to abandon the linear model the moment the incident gets complex.
How should you communicate incident status to leadership in a way that is actually useful?
Leadership needs answers to four questions: (1) What happened? (2) What is the current business impact? (3) What are we doing about it right now? (4) What do you need from us? Frame every SITREP around these four questions, not around phase names. Compare: "We are currently in the Containment phase" (meaningless) versus "We have 23 confirmed compromised systems, 18 are isolated, 5 are being investigated. Payroll processing is unaffected. We need authorization to bring in our external IR retainer at an estimated cost of $X for forensic support." The second version enables a decision. The first version creates the illusion of information. Frequency: SITREP every 2 hours during active IR, every 4 hours once containment is stable. Go silent only when there is genuinely nothing new — and even then, send a brief "no status change" update so leadership knows the team is still working, not stuck.
Why do some organizations insist on running tabletop exercises when they know a tabletop is not the same as a real incident?
Tabletops test process, communications, and decision-making — the things that fail in real incidents for reasons that have nothing to do with technical skill. A technically excellent team that has never rehearsed their escalation path will fumble it at 2 AM: who calls the CISO? What is their cell number? What do you say? Is legal counsel pre-engaged or do you need to find a firm during the incident? Does the IR retainer have a 24/7 hotline, and does anyone know the number? Tabletops expose these gaps safely. They also reveal assumption mismatches between teams — the security team assumes the network team will immediately implement a firewall block on request; the network team assumes they need a change control ticket. Those mismatches are invisible until tested, either in a tabletop or in a real incident. Run tabletops not to practice technical skills but to practice process, communication, and decision authority under pressure.