← all systems · see it in The System
Incident response
What an organisation does when a computer security incident happens: 'a violation or imminent threat of violation of computer security policies, acceptable use policies, or standard security practices' (src:nist-sp800-61r2) is detected, analysed, contained, eradicated and recovered from, and learned from, by a response team following a plan (src:wp:computer-security-incident-management). Walked as the practice and the capability that does it: the signs coming in, the triage desk, the incident handlers and their lead, the plans and procedures, the lessons learned, and the wider web of coordination centres (CERT/CC, FIRST) that grew up after the 1988 Morris worm.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Incident response and recovery |
| born at | November 1988 (CERT/CC, Pittsburgh, at DARPA's direction after the Morris worm) |
| born precision | month |
| boundary | Inside: the incident response capability of an organisation (or of a community, for a coordinating team): its policy, plan and standard operating procedures; the reporting mechanisms and the triage function that receives every sign; the incident handlers, the incident lead and the team manager; the analysis, prioritisation, notification, containment, eradication and recovery work; the evidence handling; the lessons learned, follow-up reports and incident data. Outside: the monitored systems themselves; the detection tools that raise alerts (IDPS, SIEM, antivirus) and the security operations centre that watches them around the clock (the edge is drawn differently by different teams: 'The first tier of an incident response team often assumes responsibility for intrusion detection'); the attackers and their malware; prevention and protection, which the 2025 guide places outside incident response itself (contention C-1575); the organisation's management, legal, public affairs and business continuity groups the team depends on; law enforcement, regulators, media and peer teams. |
| nested level | A practice of computer and information security (rung 12), part of an organisation; its handlers are people (Human), its tools run on computers and networks, and its teams are linked across organisations by coordination centres and forums (CERT/CC 1988, FIRST 1989). |
| scale | social |
Measures (21) published science
| ransomware attacks on US health care delivery organisations | 43 (2016) to 91 (2021); 374 in all attacks per year |
| escalation wait before raising a level | perhaps 15 minutes |
| lessons learned meeting after a major incident | within several days |
| breach report deadline (Netherlands) | 72 hours after discovery |
| recovery time of large-scale incidents | months (early phases days to weeks) time |
| incident response duration, 2012 picture against 2025 | usually within a day or two (then); weeks or months (now) time |
| dwell time before detection (SolarWinds) | 14 months |
| average time to detect and contain a data breach (IBM 2023, as cited) | 204 days |
| share of alarms analysts call false positives | 99 % (one analyst's figure) |
| share of organisations struggling with EDR alert volume | nearly 70 % |
| Internet hosts at the Morris worm | 60,000 (estimate); more than 36 million a decade later; 171.6 million in January 2003 hosts |
| machines infected by the Morris worm | around 6,000 (usually reported; a guess at 10% of 60,000) machines |
| share of health care ransomware attacks that disrupted care | 44.4 (166 of 374) % |
| share of ransomware victims restoring from backups | 20.6 % |
| health care ransomware attacks reported late to HHS | 203 (54.3% of all attacks), outside the 60-day window |
| CrowdStrike outage at one health system: devices crashed | 8647 BSOD events; 729 devices needing manual remediation |
| command center open (CrowdStrike outage, one health system) | approximately 30 hours |
| share of CryptoLocker victims who paid | up to 40 (about two-thirds of payers recovered files) % |
| major errors attributed to handover miscommunication (health care, cited) | 80 % |
| WannaCry: start to kill-switch halt | 07:45 to 15:03 UTC, 12 May 2017 time of day (UTC) |
| FIRST membership | 810+ organisations from more than 110 countries |
Open contentions (5) open contention
Where sources or readings disagree. Both stay on the map.
- C-1496 Are 99% of security alarms really false?
booked: The map books the analysts' own count, and keeps the finding that most such alarms are real events on harmless activity.
the other reading: most alarms are benign triggers, true but harmless; false alarms in the strict sense dominate - C-1575 Where does incident response start and end?
booked: The map books the NIST outline, and keeps the other published outlines.
the other reading: incident response is a circular life cycle of a separate team that includes preparation (SP 800-61r2, 2012); incident response is integrated across all six CSF Functions and organisational operations, with continuous improvement (SP 800-61r3, 2025); incident response is only one component of a broader incident handling service (CSIRT handbook, 2003) - C-1576 Is incident response one system with the body's immune response?
booked: The map keeps both readings: one job that starts with life, or a shared pattern in separate systems.
the other reading: answering damaging events is one system that begins with life (immune response) and computer incident response is its latest kind; computer incident response is a kind of emergency management (rung 11), with fire services and ICS as elder siblings; they are different systems that share a job - C-1581 Was the Morris worm released on 2 or 3 November 1988?
booked: The map keeps both dates.
the other reading: 2 November 1988; 3 November 1988 (the CERT article's date) - C-1582 How many machines did the Morris worm infect?
booked: The map keeps the usual figure of about 6,000 and the account that it was a guess.
the other reading: about 6,000 machines infected; the figure was made up as ten percent of an estimated 60,000 Internet computers
Cross-scale patterns seen here (21) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Incident response. Each links to every other system that shows it.
- borrowed-then-held (structural)
Both phases shown: after the 1988 worm one centre, the CERT/CC, answered for the whole Internet; as the network grew 'a single CSIRT is unable to effectively serve such a vast constituency', and hundreds of organisations formed their own teams. Inside an organisation the same move runs from outsourc - lightning (structural)
A sudden discharge that changes what follows: the Morris worm of 2 November 1988, replicating regardless of infection status, ran away across the young Internet; within weeks DARPA had a CERT/CC in Pittsburgh, and the WANK worm next year led teams to form FIRST. Each major incident in the record (Wa - same role, different builds (structural)
One job, answering a damaging event as it happens (take the call, size it up, contain it, put it out, restore, and keep a record for next time), done by unrelated builds: the computer security incident response team; the fire department the CSIRT handbook compares it to; the Incident Command System - templating (structural)
The same module reused unchanged across teams and still working: the NIST life cycle and its SOP elements, the CSIRT handbook's triage-handling-announcement-feedback functions, tracking-number schemes (CERT#, VU#), and the Incident Command System, which 'has acted as a pattern for similar approaches - The carry (structural)
Overflow passed on and the wait, both faces: work starts with one handler who must decide whether to hand the incident to a more experienced colleague or call in others; an unanswered call waits perhaps 15 minutes and then is escalated a level, again until someone answers; an incident too big for it - the whole sets the local reference (structural)
Each organisation's own policy defines what counts as an incident, so each team's count reads true locally, yet the counts do not close into one shared frame: 'what one organization considers a single incident may be considered multiple incidents by others', and sensors label port scans by different - Umwelt: each system carves its own slice of one shared medium (structural)
One shared world of events, carved differently: from the target's side most attacks show no precursor; an outside MSSP sees across many customers and spots new threats sooner, while in-house staff know the environment and can tell the organisation's own odd behaviour from attack; US-CERT sees trends - who opens the door (structural)
Every way in is a port opened from inside ahead of time: an advertised single point of contact with phone, email, forms and at least one anonymous route; preregistered contacts and tracking numbers; relationships with law enforcement made 'before an incident occurs'; information-sharing rules agreed - causal ceiling (partial)
A limit on what the target can know, set by where it stands: most attacks have no detectable precursor 'from the target's perspective'; the part seen at onset is the tip of an iceberg; trends invisible to one organisation appear only across many. Partial: the sources do not separate the structural l - holding / entrainment (partial)
The organisation sets the clock of its team: shift schedules and handovers, follow-the-sun passing between time zones, the plan reviewed at least annually, the lessons learned meeting within several days. Partial: the clock is set by policy, not by a held loop the sources describe. - outside-to-inside (partial)
The watch moves from the boundary inward: insider incidents prompt 'stronger security controls on internal networks (e.g., deploying intrusion detection software to more internal networks and hosts)'. Partial: the outside phase (perimeter watch) is implied, not dated in the source. - red-queen (partial)
Attackers subtly modify attacks to slip rules ('nearly half of the current SIEM rules could be easily evaded'); teams 'should evolve to reflect new threats'; incidents have grown more frequent and damaging. Partial: both sides improving is shown; the stalemate (neither gaining ground) is not measure - runaway from an unstable balance (partial)
Below capacity the team keeps up; past it, more incidents delay recovery, which lets damage spread and makes more work: 'This could overwhelm the resources and capacity for response, which would result in delayed or incomplete recovery and possibly more extensive damage'. Partial: the balance point - same shape, opposite sign (partial)
One act, cutting a host off the network, with its effect's sign reversed: normally it stops the spread; with malware that pings a partner host, the failed ping sets off overwriting or encryption of the drive. Partial: the shared process is the disconnection; the reversal depends on the malware's tri - The Payment to Entropy (partial)
Every hand-off pays a toll in signal: millions of signs a day reduce to a handful of incidents, and one analyst puts the unusable share of alarms at 99% (usable out per alarm sent in about 1 in 100); at each handoff between people, shifts or teams information is lost ('Cyber incident handling errors - trade one property for another (partial)
Containment buys safety with service and evidence: its criteria weigh damage against 'Need for evidence preservation' and 'Service availability'; a sandbox buys evidence with risk and liability; chasing the attacker's identity costs the primary goal. Strategies settle at different points (emergency - cephalization (surface)
Intake gathers at the end that meets new input first: one advertised focal point, the hotliner who reads all mail and answers all calls. Surface: no source gives the input-first reason as a selection pressure. - nested ladder (surface)
Ordered scales where each rung is a stronger form of the last: recoverability Regular, Supplemented, Extended, Not Recoverable; functional impact None to High; and the escalation ladder. Surface: they are ordered ratings, not shown to imply the rungs below. - partner-becomes-part (surface)
Outside specialists, contractors and clinical informaticists join the team for an incident; an onsite contractor can be the team. Surface: the partners keep their own identity (the dark twin: separate inheritance). - same toolkit, independent builds (surface)
Fire services (ICS, from 1970s wildfires) and computer incident response (CERT/CC, 1988) assembled the same parts (one number to call, an incident commander, escalation, after-action learning). Surface: the builds are not independent; the CSIRT handbook draws on the fire analogy and NIMS frames both - symmetry breaking (name the cut) (surface)
The cut that starts everything: among events alike in kind, one is judged an incident ('a matter of judgment'), and from then the response runs one way. Surface: a cut is named but there is no symmetric dynamics before it.
Valence: what it trades, and with whom (26) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Security operations center (SOC) via sys:electromagnetic-field · Incident response gain
- Human (Homo sapiens) via sys:human-society · Incident response gain
- Government via sys:human-society · Incident response mixed
- Law (rules and law) via sys:human-society · Incident response mixed
- Mass media via sys:human-society · Incident response mixed
- Operating system via sys:computer-hardware · Incident response gain
- Police via sys:human-society · Incident response mixed
- Health care via sys:written-record · Incident response mixed
- Computer security via sys:electromagnetic-field · Incident response gain
- Malware via sys:operating-system · Incident response loss
- Intrusion detection system via sys:electromagnetic-field · Incident response gain
- Threat actor via sys:human-society · Incident response loss
- Information security via sys:written-record · Incident response gain
- Organization via sys:human-society · Incident response gain
- Computer network via sys:electromagnetic-field · Incident response gain
- Supply chain security via sys:software · Incident response loss
- CERT Coordination Center via sys:written-record · Incident response gain
- Outsourced incident response (MSSP, contractors) via sys:electromagnetic-field · Incident response mixed
- FIRST (Forum of Incident Response and Security Teams) via sys:written-record · Incident response gain
- Security policy and governance via sys:incident-response · Incident response gain
- Risk management via sys:written-record · Incident response gain
- Security information and event management (SIEM) via sys:written-record · Incident response gain
- Cyber threat intelligence via sys:written-record · Incident response gain
- Digital forensics via sys:written-record · Incident response gain
- Vulnerability management via sys:organization · Incident response gain
- Business continuity via sys:written-record · Incident response gain
Partners and parts published science
Built from / builds (13)
- Computer security
- Human (Homo sapiens)
- Incident response policy, plan and procedures
- Incident response team (CSIRT)
- Triage function (single point of contact)
- Incident lead / incident commander
- Issue tracking system and incident records
- Digital forensics
- Lessons learned meeting and follow-up report
- Information security
- Slammer against human response (25 January 2003)
- Ransomware extortion of victims
- Attackers against incident response
Sits inside / holds (5)
expressed by (25)
- Generation/Synthesis
- Templating / Standardization
- Quality Improvement/Refinement
- Interoperability/Transduction (Bridging)
- Symbiosis/Co-Creation (Collaborative Specialization)
- Distributed Autonomy/Decentralized Execution (Subsidiarity)
- Internal Coordination/Regulation
- Sensing/Scanning
- Uncertainty Reduction/Learning
- Multi-Scale Data Fusion / Triangulation
- Trail-Marking / Externalized Memory (Stigmergy)
- Offline Simulation & Replay-Guided Planning
- Social Learning / Public Information Use
- Quick Harm Avoidance
- Reactive Response
- Proactive Defense
- Self-Monitoring
- Delay/Redirection Tactics
- Validation of Identity/Intention
- Validation of Effectiveness
- Enclave/Segregation
- Resilience & Repair
- Redundancy & Diversity
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (23)
- Computer security
- Security operations center (SOC)
- Human (Homo sapiens)
- Law (rules and law)
- Operating system
- Intrusion detection system
- Threat actor
- Organization
- Computer network
- Security information and event management (SIEM)
- Antivirus / anti-malware software
- Incident response policy, plan and procedures
- Incident response team (CSIRT)
- Triage function (single point of contact)
- Incident lead / incident commander
- Issue tracking system and incident records
- Digital forensics
- Lessons learned meeting and follow-up report
- Incident Command System (emergency services)
- IT help desk
- Information security
- Cyber threat intelligence
- Business continuity
Kinds (6)
- Central incident response team
- Distributed incident response teams
- Coordinating team (a CSIRT for CSIRTs)
- National CERT / CSIRT
- Outsourced incident response (MSSP, contractors)
- Virtual (part-time, ad hoc) team
wired (3)
Notes published science
- The famous count of 6,000 machines hit by the 1988 Morris worm may never have been counted at all: Paul Graham says 'someone guessed that there were about 60,000 computers attached to the Internet, and that the worm might have infected ten percent of them' (src:wp:morris-worm). The worm still led DARPA to set up the first incident response team, the CERT/CC, that same November (src:wp:cert-coordination-center).
- Pulling the plug can be the trigger: NIST warns that a compromised host may ping another host, and when responders disconnect it 'the subsequent pings will fail' and the malware 'may overwrite or encrypt all the data on the host's hard drive' (src:nist-sp800-61r2). The textbook first move of containment is, against some malware, exactly what sets it off.
- Unlike fire crews, cyber responders start alone: the work 'is mostly individual until a non-routine or unfamiliar incident occurs', and the first person has to decide when to call for help, while in firefighting 'tasks usually begin at the team level' (src:vanderkleij-2017-csirt). Under time pressure people wrongly decide an incident is familiar and don't ask.
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
incident response as incident handlers describe it: signs of an incident in the organisation's systems up to the mission it protects (reader's view)
- L1: The organisation's systems, networks, data and users, where events happen; an event is any observable occurrence, an incident a violation or imminent threat of violation of policy
- L2: Signs of incidents gathered: precursors and indicators from IDPSs, SIEMs, antivirus, file integrity checks, logs, flows, third-party services and people
- L3: Triage: a single point of contact accepts, acknowledges, tags, tracks and gives an initial priority
- L4: Detection and analysis: validate, correlate, scope, prioritise
- L5: Coordination and decision: incident lead, escalation, notification, containment, eradication and recovery
- L6: Policy, plan, procedures and lessons learned
- L7: Mission: minimise loss and disruption, restore services, meet the law
the boardroom (shared floors)
- L3: Triage: one focal point accepts every report and alert, acknowledges it, gives it a tracking number, tags it to an existing or new event and assigns an initial priority (the help desk, the hotliner, or the first tier of a security operations centre)
- L4: Detection and analysis: incident handlers validate indicators, correlate events across sources, determine scope, and prioritise by functional impact, information impact and recoverability
- L5: Coordination and decision: the incident lead (incident commander) coordinates the handlers, escalates, notifies, and decides containment, eradication and recovery; handovers between shifts and teams
- L6: Policy, plan, procedures and learning: the incident response policy and plan, standard operating procedures, lessons learned meetings, follow-up reports and incident data; the retention face of sensemaking
- L7: Mission and governance: minimise loss and disruption and restore normal services; management commitment, budget and staffing; law and the organisation's risk strategy (CSF Govern)
way in: Attacker communication channels (e.g. IRC channels bots and attackers use)
- L1: Channels attackers and their bots use
- L2: Watched by handlers; what is seen is a lead, not a fact
way in: Forensic evidence from affected systems (snapshots, disk images, preserved logs)
- L1: The machine's state, changing as the incident runs
- L2: Acquired with forensic tools: disk images, packet captures
- L3: Logged and passed hand to hand under chain of custody
way in: Law, regulation and reporting duties (breach notification deadlines, mandatory reporting)
- L1: A reporting duty with a deadline
- L2: Read by the legal department into plans and procedures
way in: Logs and flows (operating system, service, application and network device logs; network flows)
- L1: What the systems record: accounts accessed, actions performed, sessions between hosts
- L2: Kept by a log retention policy and centralised
- L3: Automation selects events of interest for human review
way in: New vulnerabilities, exploits and threat advisories
- L1: Public information on new vulnerabilities and exploits
- L2: Fed in automatically (NVD feeds) and tracked under their own numbers (VU#)
way in: People inside the organisation reporting signs (users, administrators, security staff)
- L1: Something a person notices (an odd filename, bounced mail, a defaced page)
- L2: Reporting mechanisms set up in advance, one of them anonymous; the help desk as first point of contact
way in: Reports from outside (other organisations, other incident response teams, coordination centres)
- L1: An outside party says something is wrong (a claim of attack, a defaced page)
- L2: Monitored mechanisms for external reports, forwarded to the help desk
way in: Security tool alerts (IDPS, SIEM, antivirus, file integrity checks)
- L1: Events on hosts and networks: the medium the tools watch
- L2: Detection tools turn events into alerts (signatures, anomaly checks, checksums)
way in: The team's own record of past incidents (tickets, knowledge base, follow-up reports)
- L1: The issue tracking system and the follow-up reports
- L2: Queried for related activity
- L3: New information tagged to existing or past events
way in: The team's own state: workload, stress, shifts and handovers
- L1: Stress, on-call duty, peaks and troughs of work
- L2: Handovers between people and shifts carry the current situation
way in: Third-party monitoring and shared indicators (blacklists, CSIRC notification lists, sector feeds)
- L1: Other organisations' incident activity linked to this organisation's addresses and names
- L2: Subscriptions and notification lists, some open only to incident response teams
way in: Threats and demands from attackers (announced attacks, extortion, ransom notes)
- L1: An announced attack: a precursor
- L2: Extortion: pay or the data is released
way out: Changes to controls, policies and procedures (lessons learned put back into risk assessment)
- L1: Controls and procedures changed
- L6: Incident data fed back into risk assessment
way out: Containment (disconnect, shut down, segment, disable functions, redirect to a sandbox)
- L1: Isolation on the medium: from a locked server room to a segmented network
- L3: Network controls altered (firewall rulesets) by information security staff
- L5: Choose a containment strategy for the incident type
way out: Eradication and recovery (remove malware, disable accounts, restore from clean backups, rebuild, patch, change passwords)
- L1: Systems restored to normal operation
- L5: Phased remediation, prioritised
way out: Follow-up report, chronology and retained evidence
- L1: A formal chronology and damage estimate, kept for legal use
- L6: A follow-up report for each incident
way out: Information sharing and advisories (to peer teams, ISACs, coordination centres; advisories inside the organisation)
- L1: Advisories issued
- L6: Taking part in sharing groups
way out: Notification and status updates (CIO, legal, human resources, public affairs, coordination centres, law enforcement, affected parties, media)
- L1: Told by email, website, phone, in person, voice mail or paper notices
- L5: Who must be told, what and when
way out: Paying a ransom (a victim's option under ransomware)
- L1: The ransom paid
- L5: Pay or lose the files