Chapter 02

Preparation

Preparation is the only IR phase where you are not improvising under pressure. Every hour of preparation time reduces three hours of incident response time — but most organizations discover their preparation gaps at the worst possible moment.

Scenario

Ransomware hits at 11 PM on a Friday. The on-call analyst escalates to the IR lead, who opens the IR runbook. The runbook was last updated 34 months ago. It references a SIEM that was decommissioned 18 months ago, an IR retainer with a firm the company stopped using 2 years back, and a communication tree that includes three people who no longer work there. The first 90 minutes of the incident are spent figuring out what tools are actually deployed, who the actual contacts are, and what the current escalation process looks like — 90 minutes during which the ransomware is still encrypting files on 40 additional hosts. Everything that happened in those 90 minutes was a preparation failure. None of it happened during the incident.

What Preparation Actually Covers

Preparation is not a one-time project. It is an ongoing operational discipline that decays continuously and requires regular investment to maintain. The gap between teams that respond effectively and teams that don't is almost always a preparation gap, not a skill gap. Smart, technically capable analysts fail during incidents because they are discovering the state of their own environment for the first time while trying to respond to an active attacker.

  Preparation — Six Pillars
  ══════════════════════════════════════════════════════════════════════

  1. TOOLING & VISIBILITY
     ─────────────────────────────────────────────────────────────────
     EDR on every endpoint. Sysmon on every Windows host without EDR.
     Centralized log collection (SIEM or equivalent). NDR on key
     network segments. Cloud-native logging (CloudTrail, Azure Monitor,
     GCP Cloud Audit Logs) retained for 90+ days.
     Test: Can you answer "what process spawned X" for any endpoint?
     If no: you have a logging gap that will hurt you in an incident.

  2. RUNBOOKS & PLAYBOOKS
     ─────────────────────────────────────────────────────────────────
     Written, tested, current. One per major scenario:
     ransomware, BEC, active adversary, cloud breach, insider threat.
     Each runbook: step-by-step with decision gates, tool-specific,
     current as of last quarter.
     Test: Can a new analyst follow the runbook without asking anyone?
     If no: the runbook is documentation, not a usable procedure.

  3. CONTACTS & COMMUNICATION TREE
     ─────────────────────────────────────────────────────────────────
     Verified phone numbers for: CISO, GC, CEO, IR retainer 24/7 line,
     cyber insurance broker, law enforcement liaison, key vendors.
     Out-of-band channel established (Signal group, separate email).
     Test: Call the IR retainer hotline right now. Does someone answer?
     If no: find out before you need them at 2 AM.

  4. LEGAL & REGULATORY PRE-ENGAGEMENT
     ─────────────────────────────────────────────────────────────────
     Legal counsel briefed on IR process and notification obligations.
     Outside breach counsel engaged (not just general corporate counsel).
     Insurance policy reviewed: what is covered? What are the
     notification obligations to the insurer?
     Regulatory reporting requirements documented: GDPR 72-hour window,
     SEC 4-day material incident rule, state breach notification laws.
     Test: Can your legal counsel tell you your notification deadline
     for a breach of customer PII? If no: they aren't ready either.

  5. TABLETOP EXERCISES
     ─────────────────────────────────────────────────────────────────
     Quarterly scenarios. Multi-team participation (security, IT, legal,
     comms, business units). Inject-based, not scripted.
     Measure: process gaps found and remediated, not "did we pass."
     Test: What assumption mismatches were discovered in the last tabletop?
     If none: the tabletop wasn't realistic enough.

  6. ASSET INVENTORY & CRITICALITY TIERING
     ─────────────────────────────────────────────────────────────────
     Every system in the environment: hostname, IP, owner, criticality,
     data classification, EDR status, backup status.
     Tier 0: domain controllers, PKI, PAM, backup systems.
     Tier 1: business-critical applications, data repositories.
     Tier 2: general servers.
     Tier 3: user workstations.
     Test: If you had to contain all Tier 0 systems simultaneously
     right now, could you do it in 30 minutes? If no: your IR will
     be slower than an attacker who already knows your environment.

The IR Retainer: What It Is and Why You Need One Before the Incident

An IR retainer is a pre-negotiated agreement with an external incident response firm that guarantees a defined response time and scope of engagement when you call. It is not a reactive contract you sign during an incident — by then, you are at the end of a queue of other organizations also calling in crisis mode, paying emergency rates, and working with a firm that knows nothing about your environment.

ElementWhat to demandWhy it matters
Response SLA First human contact within 1 hour, 24/7/365. Dedicated IR team on-site or remote within 4–8 hours for P1 incidents. Without an SLA, "we'll call you back" is the answer at 2 AM on a Sunday. Hours of attacker dwell time during the response delay.
Pre-incident engagement Environment walk-through before any incident. Firm should know your SIEM, your EDR, your AD structure, your critical assets. A firm that walks in blind during the incident spends the first 4 hours asking questions you don't have time to answer. Pre-engagement eliminates this.
Forensic toolkit pre-positioning Firm's forensic tools pre-approved for deployment in your environment. Legal counsel pre-approved for attorney-client privileged engagement. Approving tools during an incident adds delay. Attorney-client privilege must be established before the investigation, not after.
Threat intelligence integration Firm's threat intel on current campaigns and actor TTPs shared proactively, not just reactively. If the firm is tracking an active campaign targeting your industry, you want to know before you're a victim, not after.
Contract scope Clear definition of what hours are included in the retainer, what constitutes a P1 vs P2 incident, billing model for over-retainer usage. Surprise invoices during or after an incident are a distraction at the worst time.
Common mistake: treating the IR retainer as an emergency service

Many organizations sign an IR retainer and then put the number in a drawer until needed. This is a mistake for three reasons. First, you haven't verified the SLA works — the first test of whether the firm answers at 2 AM should not be a real incident. Second, the firm doesn't know your environment — the pre-incident walk-through is not just a courtesy, it is the mechanism by which the firm's ability to help you is 3–4x faster when the incident happens. Third, you haven't verified that your legal counsel has established the appropriate privilege relationship with the firm. Call the retainer line quarterly. Schedule the environment walkthrough annually. Treat the retainer as a relationship, not a fire extinguisher.

Tabletop Exercise Design

A tabletop exercise is a structured discussion-based simulation of an incident scenario. Done well, it exposes gaps in process, communication, and decision authority that are invisible until tested. Done poorly, it is a box-checking exercise that produces a completion certificate and no improvements.

  Tabletop Exercise Structure
  ══════════════════════════════════════════════════════════════════════

  BEFORE THE EXERCISE
  ─────────────────────────────────────────────────────────────────
  • Select scenario: realistic for your threat model, specific enough
    to force real decisions (not "a breach happened, what do you do?")
  • Invite the right people: not just security. Include IT, legal,
    comms, business unit leads, executives. The gaps are always at
    the interfaces between teams.
  • Brief participants: scenario is confidential until the session.
    People should not be able to "prepare" the right answers.
  • Prepare injects: unexpected developments that force pivots.

  DURING THE EXERCISE
  ─────────────────────────────────────────────────────────────────
  Facilitator presents scenario in stages. After each stage:
  • What do you do right now?
  • Who makes that decision?
  • Who do you notify?
  • What information do you need that you don't have?

  Injects (examples for a ransomware scenario):
  • "The CFO's laptop is also encrypted — she's trying to use it
    to continue the board call. What do you do?"
  • "A journalist has called the PR team. They want a statement
    in the next 30 minutes."
  • "The IR retainer firm says their earliest availability is 6 hours."
  • "The attacker has posted a sample of your customer data on their
    leak site."

  Each inject should surface a process gap or test a decision boundary.

  AFTER THE EXERCISE
  ─────────────────────────────────────────────────────────────────
  Hot wash: immediate debrief. What went wrong? What was unclear?
  Gap register: every gap identified → owner + deadline.
  Follow-up: tabletop findings tracked like any other project work.
  Measure: not "did we pass" but "how many gaps did we find and fix?"
Why the most valuable tabletop findings are about communication, not technical gaps

Technical team members expect to find technical gaps in tabletops — missing tool coverage, incomplete runbooks. The most operationally significant findings are almost always communication and authority gaps that nobody anticipated. Common examples: the security team assumes they can contain any host immediately; the IT team has a change control process that requires 24-hour approval. The IR lead assumes they can authorize a $200k external IR engagement; the CFO assumed that decision requires board approval. The security team plans to move to Signal for OOB communications; half the participants don't have Signal installed and don't know their own phone numbers to share. These gaps are not technical. They are organizational. And they will cause exactly the same delays and confusion in a real incident unless found and fixed in an exercise first.

Runbook Writing Methodology

A runbook is a step-by-step procedure for responding to a specific incident type. It is not a strategic document — it is an operational artifact designed to be followed by an analyst under pressure who may not have deep experience with the specific scenario. The test of a runbook is not "does it capture our process?" but "can someone follow it at 2 AM without asking for help?"

Runbook propertyWhat it meansHow to verify
Tool-specific Every step names the actual tool used in your environment. Not "check the SIEM" — "in CrowdStrike Falcon, go to Investigate → Event Search, filter by..." Have someone unfamiliar with the tool follow the runbook. If they ask "where is that in the UI?", the runbook fails.
Decision-gated Every decision point in the runbook explicitly branches. "If X: do Y. If Z: do W. If unclear: escalate to IR lead." Run through the runbook with a scenario where the answer is ambiguous. Does the runbook tell the analyst what to do?
Contact-complete Every person who needs to be notified is named with a direct phone number. Not "notify CISO" — "call [name] at [number]". Verify every phone number. Verify every person is still in the role.
Freshness-dated Every runbook has a "last verified" date. Any runbook older than 90 days without a verification is considered untrusted. Check: does someone own the review schedule? Is the schedule being followed?
Environment-specific Written by someone who knows your environment. Not copied from a template and lightly edited. Does it reference your actual SIEM, your actual EDR, your actual network segments, your actual critical assets by name?

Logging Gaps: The Silent Preparation Failure

Every investigation is bounded by the logs that exist. An attacker can spend months in an environment, and if their activity touches only systems with no logging, the investigation cannot tell you what they did, where they went, or what they accessed. Logging gaps are a preparation failure that only becomes visible during an incident — when it is too late to fix.

  Critical Log Sources and Their Failure Modes
  ══════════════════════════════════════════════════════════════════════

  LOG SOURCE          WHAT IT TELLS YOU          COMMON GAP
  ────────────────    ────────────────────────   ────────────────────
  Windows Security    Logons, account changes,   Event log size too
  Event Log           process creation (4688)    small → overwrites.
                      (if enabled), privilege    Process creation
                      use, policy changes        logging often disabled.

  Sysmon              Process creation with       Not deployed. Or
                      command line, network       deployed with default
                      connections, file creates,  config that misses
                      registry, process access    most value.

  EDR Telemetry       Process trees, memory       Gaps in coverage:
                      events, file events,        servers, OT/ICS,
                      network events              legacy systems, VDI.

  Authentication      Domain logons (4624/4625),  Short retention
  (AD / DC logs)      Kerberos TGT/TGS requests  (7 days default).
                      (4768/4769/4771)            Often not forwarded
                                                  to SIEM.

  Network Flow        East-west movement,         Not collected on
  (NetFlow/VPC Flow)  beaconing, data volume      internal segments
                                                  — only perimeter.

  Cloud Audit Logs    API calls, IAM changes,     CloudTrail not
  (CloudTrail etc.)   data access, config changes enabled in all
                                                  accounts. Short
                                                  retention (90 days
                                                  default for S3).

  MINIMUM VIABLE LOGGING FOR IR:
  ─────────────────────────────────────────────────────────────────
  • Sysmon on all Windows endpoints (SwiftOnSecurity or Olaf config)
  • EDR on all endpoints with 90-day telemetry retention
  • DC security logs forwarded to SIEM, 1-year retention
  • CloudTrail/Azure Monitor in all accounts, 1-year retention
  • DNS query logs (passive DNS) — lateral movement leaves DNS trace
  • Firewall/proxy logs — outbound connection visibility
Mental model: you can only investigate what you logged

Every logging gap is a blind spot that an attacker can operate in invisibly. When an investigation completes and you know what the attacker did, everything you know came from log data. The attacker's actions that you cannot account for are bounded by the logs that don't exist. Organizations that under-invest in logging do not have cleaner incidents — they have incomplete pictures of their incidents. The attacker was still there, doing things, in the places your logs don't cover. You just don't know what they did there. Preparation means eliminating those blind spots before the incident, not discovering them during.

The legal dimension of incident response is often the domain that security teams are least prepared for, and the one where being unprepared has the most direct consequences. Regulatory notification deadlines are measured in hours and days, not weeks. Outside counsel relationships must be established before the incident for attorney-client privilege to apply to the investigation. Insurance notification obligations may have tight windows that trigger coverage issues if missed.

Regulatory frameworkNotification deadlineTrigger
GDPR (EU) 72 hours to supervisory authority Breach of personal data of EU residents, unless unlikely to result in risk to individuals
SEC (US public companies) 4 business days after "materiality" determination Cybersecurity incident determined to be material (Form 8-K Item 1.05)
HIPAA (US healthcare) 60 days to HHS; 60 days to affected individuals Breach of unsecured PHI affecting 500+ individuals (or less, with state law variations)
State breach notification (US) Varies: 30–90 days, some as low as "expedient" PII of state residents accessed or acquired without authorization
Cyber insurance Typically 24–72 hours Confirmed or suspected security incident — check your specific policy
Common mistake: not establishing attorney-client privilege before the investigation

Attorney-client privilege protects communications between a client and their attorney from compelled disclosure in litigation. For this privilege to cover an incident investigation, outside counsel must be engaged before the investigation begins — ideally, the investigation is conducted at the direction of counsel. If your internal security team conducts the investigation independently and outside counsel is only engaged afterward, the investigation documents may not be privileged. This means the forensic findings, the IR report, the timeline, and all internal communications about the incident may be discoverable by opposing parties in litigation. The correct sequence: when you confirm a significant incident, call outside breach counsel immediately, before the investigation proceeds. Have them engage the IR firm on your behalf. Every investigation document is then created under legal direction and potentially privileged.

Q & A

How often should IR runbooks be reviewed and updated?

Every time a relevant tool, process, or personnel change occurs — and at minimum every 90 days as a scheduled review. Changes that require immediate runbook updates: SIEM migration, EDR platform change, any tool referenced in the runbook being decommissioned, any person named in the runbook leaving the organization, significant network architecture changes, and any lessons-learned item from an actual incident that reveals a runbook gap. The review should not just verify accuracy — it should test usability. Have someone who didn't write the runbook attempt to follow it against a realistic scenario. The gaps they encounter are the gaps that will cost you in a real incident.

What makes a tabletop exercise realistic enough to be valuable?

Three things: the scenario must be plausible for your specific threat model (not a generic "breach happened" scenario but "a phishing email targeting finance users delivered a payload that beacons to a cloud storage service" — specific enough to force real decisions about your real tools and processes), the injects must be unexpected and must force pivots that reveal assumption mismatches, and the right people must be in the room. A tabletop attended only by security team members is nearly useless — the gaps are always at the interfaces with other functions (IT, legal, comms, business). The most valuable tabletops routinely discover that the security team assumed they had decision authority they don't have, or assumed they could use communication channels that aren't actually available in an incident. These discoveries are cheap in a tabletop. They are very expensive during a real incident.

The company doesn't want to pay for an IR retainer. How do you justify the cost?

Frame it in terms of the cost of the alternative. An IR retainer costs $30k–$150k/year for a mid-market organization. An unretained emergency response engagement starts at $500/hour and typically runs 500–2,000 hours for a significant incident — $250k–$1M in fees alone, at emergency rates, with a firm that doesn't know your environment. The retainer also includes hours that can be used proactively — threat hunts, runbook reviews, tabletop facilitation — so it is not a pure insurance cost. Present it as: "The retainer costs X. Not having it costs 3–5X in the first incident, plus whatever damage occurred during the additional response time. Given that our industry is actively targeted and our last tabletop identified 14 preparation gaps, I assess this as a high-probability event, not a theoretical risk."

How do you handle the situation where key IR personnel leave the organization?

IR team continuity requires three things: documentation, cross-training, and succession. Documentation: every critical process, tool configuration, and institutional knowledge lives in written form that doesn't reside only in a person's head. When someone leaves and you discover knowledge gaps, those gaps existed before they left — you just didn't know it. Cross-training: no single person should be the only one who knows how to use a critical IR tool or execute a critical IR process. Succession: when an IR lead or key analyst gives notice, the runbook review and knowledge transfer become immediate priorities, not a future task. The specific failure to avoid: an incident occurring within 6 months of a key departure, with the new person discovering on the fly that their predecessor was the only one who knew how the SIEM worked, what the EDR console password was, or where the forensic workstation was physically located.

What should be in an asset inventory for it to be useful during an incident?

Hostname, IP address (current — not the one from last year's scan), operating system and version, primary owner/team, business criticality tier, data classification (does it store PII, credentials, IP, financial data?), EDR deployment status (covered/uncovered/exempted), backup status and last tested restore date, and network segment/VLAN. The minimum viable use case: during an incident, you receive an alert referencing a hostname. You look it up in the asset inventory and immediately know: how critical it is (determines containment priority), whether EDR is deployed (determines what forensic data is available), what data it might hold (determines notification obligations), and who owns it (determines who to contact for business impact assessment). Without this, every question requires an investigation of its own at a time when you have no spare capacity.

What logging retention periods should be the minimum standard?

The minimum standard for IR utility: 90 days for endpoint telemetry (EDR and Sysmon), 1 year for authentication logs (domain controller security logs), 1 year for cloud audit logs (CloudTrail, Azure Audit, GCP Audit), 90 days for network flow data, and 7 years for formal incident records and post-incident reports (litigation timelines mean you may be asked about an incident years later). The practical reason these minimums matter: sophisticated intrusions often have dwell times measured in months. An investigation of a breach discovered today may need to reconstruct events from 60, 90, or 120 days ago to find the initial compromise vector. If your logs only go back 30 days, that portion of the investigation is permanently inaccessible. The cost of storage is trivial compared to the cost of an incomplete investigation. Budget for retention before you need it.