← all systems · see it in The System
Cyber threat intelligence
A part of cybersecurity that collects, analyses and shares information about potential or existing cyber threats (src:wp:cyber-threat-intelligence): evidence-based knowledge about existing or emerging threats (src:mavroeidis-2021-cti-model). Walked as the practice and the trade: the intelligence cycle as security practice runs it, indicator feeds and the platforms that hold them, analysts and their frameworks (kill chain, Diamond Model, ATT&CK), naming and attribution, sharing bodies (ISACs, TLP, STIX/TAXII), the paid and open market, and the limits the research measures.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Cyber threat intelligence (trackers), sys:threat-intelligence, Threat intelligence and signature vendors |
| born at | 2013 as a named industry (APT1 report); forerunners 1948 (intelligence cycle), 1998-1999 (PDD-63, sector ISACs), early 2000s (TLP, blacklists), 2011 (kill chain) |
| born precision | years to decades (contested) |
| boundary | Inside: the work of producing threat intelligence (requirements, collection, processing, analysis, production, dissemination), the people who do it (CTI analysts in vendors, SOCs and government), their platforms (threat intelligence platforms, MISP), their frameworks and standards (kill chain, Diamond Model, ATT&CK, STIX/TAXII, TLP) and the feeds and reports they make. Outside: the threat actors it watches, the SOCs, incident responders and leaders who consume it (their use floor is booked as CTI's L6 where it nests in a SOC), sharing bodies (ISACs), governments and law enforcement. Edges contested: whether CTI is a process or only a product (C-1598). |
| nested level | A practice and an industry at rung 12, part of cybersecurity, borrowed from military and government intelligence; run inside organisations (in-house teams, SOCs) and as a market of vendors and sharing communities. |
| scale | social |
Measures (23) published science
| characteristic clock: feed pull cycle | once a day, the indicators of the last 24 hours per day |
| indicator validity window (common query window) | 30 days |
| phases of the intelligence cycle | 5 phases |
| feeds measured (Li et al.) | 47 IP feeds and 8 malware hash feeds |
| share of indicators unique to one feed | 73% (scan), 88% (brute force), over 97% (other categories) % of indicators |
| coverage of large scanners by scan feeds | about 10 % |
| indicator overlap between two leading paid vendors on 22 shared actors | 2.5 to 4.0 % |
| delay before a shared indicator appears at the second paid vendor | about one month (average) time |
| commercial threat intelligence market | over 5 billion USD (global) |
| high-end paid TI subscription | 100k to 650k USD per year |
| requests for information in a paid contract | about 10 per year |
| threat intelligence held at one SOC | 17 then 20 two months later terabytes |
| share of collected TI a SOC aims to filter out | half (per-hand-off yield about 0.5) fraction |
| sightings in a MISP community | 10,219 sightings, 30 confirmed false positives |
| threat actor names across vendor taxonomies | 3,287 names in 977 alias clusters (17 taxonomies, 13,371 reports) |
| ATT&CK at public release (May 2015) | 96 techniques under 9 tactics |
| ATT&CK tactics today | 14 tactics |
| indicator publication after a vulnerability (CVE-2023-34362) | 149 IPv4 indicators over 46 days, starting 30 days after the CVE |
| users satisfied with timeliness of intelligence | 52 (46 dissatisfied with expired indicators not removed) % of survey respondents |
| strategic intelligence forecasts scored | 1,514 (discrimination and calibration very good; underconfident) forecasts |
| epidemic intelligence: sensitivity of single systems vs pooled | 38-72% single, 93% combined |
| one campaign's intrusions studied | three intrusions over seven months |
| share of breaches through exploited vulnerabilities (IBM) | 47 % of attacks |
Open contentions (6) open contention
Where sources or readings disagree. Both stay on the map.
- C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: skilled intuition from experience; other theories of where thoughts come from (gameplans/theories-of-mind.md) - C-1118 In cybersecurity, do attackers have the advantage?
booked: The map keeps both: attack is easier because defenders must cover everything, and attackers have no inherent advantage because one block can break their chain.
the other reading: attacker advantage; no inherent advantage under intelligence-driven defence - C-1583 Who is behind the named hacking groups?
booked: The map books each attribution as stated, with its confidence, and keeps the denials and false-flag readings.
the other reading: as registered - C-1597 When did cyber threat intelligence begin?
booked: The map dates the named field to the 2013 APT1 report and keeps the earlier blacklists, sharing centres and the intelligence cycle as forerunners.
the other reading: CTI begins with the early spam and intrusion-detection blacklists; CTI is the 1948 intelligence cycle applied to computer networks; its birth is that of military intelligence's cyber branch - C-1598 Does cyber threat intelligence really run as a cycle of steps?
booked: The map books the cycle the practice describes and keeps the critique that most buyers just consume feeds with no process around them.
the other reading: CTI is a product without a process: most organisations are consumers, not customers, and feeds are consumed directly (Oosthoek and Doerr) - C-1599 Does cyber threat intelligence deliver what it promises?
booked: The map keeps both: reviews find it helps, and measurements find feeds overlap little, cover little and arrive late.
the other reading: CTI delivers what it promises; CTI is 'a broken product due to a broken process'
Partners and parts published science
Built from / builds (14)
- Security operations center (SOC)
- Traffic Light Protocol
- MITRE ATT&CK
- Diamond Model of Intrusion Analysis
- Threat intelligence analysts
- Threat intelligence platform (TIP)
- Indicators of compromise
- Threat intelligence feeds
- Intelligence cycle
- Cyber kill chain
- STIX and TAXII
- Threat intelligence vendors
- Actors against trackers for the actor's anonymity
- Adversaries against CTI pipelines for the integrity of shared knowledge
Sits inside / holds (3)
Trades with (valence) (21)
- Signals intelligence via sys:intelligence-information · Cyber threat intelligence gain
- Intrusion detection system via sys:written-record · Cyber threat intelligence gain
- Security operations center (SOC) via sys:written-record · Cyber threat intelligence gain
- Threat actor via sys:written-record · Cyber threat intelligence mixed
- Risk management via sys:threat-actor · Cyber threat intelligence gain
- Security information and event management (SIEM) via sys:written-record · Cyber threat intelligence gain
- Cyberwarfare via sys:internet · Cyber threat intelligence gain
- Cybercrime and its underground economy via sys:internet · Cyber threat intelligence gain
- Human (Homo sapiens) via sys:mental-representation · Cyber threat intelligence gain
- Government via sys:written-record · Cyber threat intelligence gain
- Large language model via sys:written-record · Cyber threat intelligence mixed
- Police via sys:human-society · Cyber threat intelligence even
- Malware via sys:malware · Cyber threat intelligence gain
- Incident response via sys:written-record · Cyber threat intelligence gain
- Botnet via sys:written-record · Cyber threat intelligence gain
- Organization via sys:human-society · Cyber threat intelligence gain
- Computer network via sys:computer-network · Cyber threat intelligence gain
- Underground forums via sys:underground-forum · Cyber threat intelligence gain
- Outsourced incident response (MSSP, contractors) via sys:human-society · Cyber threat intelligence mixed
- Information Sharing and Analysis Center (ISAC) via sys:human-society · Cyber threat intelligence gain
- Vulnerability management via sys:internet · Cyber threat intelligence gain
expressed by (23)
- Generation/Synthesis
- Templating / Standardization
- Quality Improvement/Refinement
- Interoperability/Transduction (Bridging)
- Amplification/Scale-Up (Manufacture & Distribution)
- Symbiosis/Co-Creation (Collaborative Specialization)
- Sensing/Scanning
- Hypothesis/Goal Selection
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Multi-Scale Data Fusion / Triangulation
- Trail-Marking / Externalized Memory (Stigmergy)
- Frontier Outposts / Sensing Extension
- Hypothesis Portfolio / Multiple Hypothesis Tracking
- Social Learning / Public Information Use
- Reactive Response
- Proactive Defense
- Technique-Based Identification (a.k.a. Technique-Based Detection)
- Validation of Identity/Intention
- Validation of Effectiveness
- Redundancy & Diversity
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (24)
- Security operations center (SOC)
- Risk management
- Cyberwarfare
- Human (Homo sapiens)
- Large language model
- Intrusion detection system
- Incident response
- Threat actor
- Traffic Light Protocol
- Organization
- Computer network
- Security information and event management (SIEM)
- Honeypots (decoys)
- MITRE ATT&CK
- Diamond Model of Intrusion Analysis
- Underground forums
- Outsourced incident response (MSSP, contractors)
- Information Sharing and Analysis Center (ISAC)
- Threat intelligence analysts
- Threat intelligence platform (TIP)
- Threat intelligence feeds
- Cyber kill chain
- STIX and TAXII
- Threat intelligence vendors
Kinds (7)
- Tactical threat intelligence
- Operational threat intelligence
- Strategic threat intelligence
- Technical threat intelligence
- Paid (commercial) threat intelligence
- Open threat intelligence
- Shared threat intelligence
Notes published science
- Two market-leading paid threat-intelligence vendors, both tracking the same 22 hacker groups, shared only 2.5-4% of their indicators, and the few they shared reached the second vendor about a month later (src:bouwman-2020-paid-ti).
- Birds and monkeys that listen to other species' alarm calls face the same arithmetic as companies sharing threat intelligence: eavesdropping pays most when the caller faces the same predators and your own detection is poor, and the most relevant calls turn out to be the less reliable ones overall (src:turner-2023-eavesdropping).
- One Russian group goes by at least four vendor names (Fancy Bear, APT-28, Sofacy, STRONTIUM, and more); across 17 vendor taxonomies, 3,287 actor names collapse into 977 clusters (src:oosthoek-doerr-2021, src:roa-2025-hesperus).
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.
cyber threat intelligence as CTI researchers and analysts describe it: adversaries' activity and its traces up to the decisions it is meant to direct (reader's view)
- L1: The threat landscape: adversaries' activity and the traces it leaves (IP addresses, domains, file hashes, malware, tactics, talk on forums), seen in many places
- L2: Collection: internal telemetry, vendor feeds, open sources, dark-web monitoring, malware analysis, government and community reports
- L3: Processing: raw information filtered and prepared (decryption, translation, data reduction), normalised into machine-readable formats (STIX), scored and aged
- L4: Analysis: indicators linked into intrusions, campaigns and actors with the kill chain, the Diamond Model and ATT&CK; attribution with stated confidence
- L5: Production and dissemination: tactical indicators, operational reports on actors and campaigns, strategic briefings; feeds, reports and TLP-labelled sharing
- L6: Use: the SOC enriches alerts and blocks infrastructure, incident responders and hunters follow the leads, leaders set priorities
- L7: Planning and direction: the customer of the intelligence states what it needs to know; the aim is less danger and less uncertainty for the decisions it serves
the boardroom (shared floors)
- L4: Analysis: analysts and platforms link indicators into intrusions, campaigns and actors (kill chain phases, the Diamond Model's adversary, capability, infrastructure and victim, ATT&CK techniques); attribution with low, medium or high confidence
- L5: Production and dissemination at three levels: tactical (indicators for SIEM rules, firewalls, endpoint tools), operational (actors, TTPs, campaigns), strategic (risk for executives and boards); delivered as feeds, reports and briefings, labelled for sharing
- L6: Use: the intelligence becomes action in the SOC (alert enrichment, blocking), in incident response and threat hunting, and in leaders' priorities; information that leads to no action is not intelligence
- L7: Planning and direction: the customer's requirements set what is collected and analysed, to give leadership relevant intelligence that reduces danger and uncertainty
way in: Analysts' tacit knowledge and judgement (inner)
- L1: Tacit knowledge, hard to put into algorithms
- L2: Judgements arrive with a felt confidence; strategic analysts discriminate well but are underconfident
- L3: Shaped by the craft's tools: priority intelligence requirements, Kent's words of estimative probability
way in: Dark web and underground forum monitoring
- L1: Data from the dark web and deep web
- L2: Dark web monitoring as a collection source
way in: Government indicators and advisories
- L1: Classified and declassified indicators held by government, shared by law since 2015 (US)
- L2: Released to companies under the 2015 act
way in: Indicator feeds (open lists, commercial vendors, aggregators)
- L1: Indicators of compromise published by others: simple observable behaviours that signal compromise
- L2: Subscribed feeds, pulled on a cycle (paid, open and shared); organisations subscribe to many
- L3: Held for a validity window set by how the user queries the feed (30 days is common), then expired
way in: Internal telemetry (logs, alerts, endpoint data from the organisation's own estate)
- L1: Activity on the organisation's own networks and endpoints
- L2: Security tools record it as telemetry and logs; internal and external data must be combined
- L3: The threat intelligence platform aggregates it with outside indicators and correlates them
way in: Load on the analysts (volume, notification fatigue)
- L1: Volume: one SOC held 17 then 20 terabytes of threat intelligence two months apart
- L2: Felt as too much to examine
- L3: Shows as filtering and ignoring a frequent source
way in: Malware samples and intrusion artefacts
- L1: Malware and artefacts recovered from intrusions
- L2: Analysis reveals what intrusions share (the same encryption key, the same shellcode)
- L3: Computed indicators: hash values and regular expressions
way in: Open sources (vendor and government reports, news, social media, vulnerability notices)
- L1: Published threat reports: security vendors the biggest producers, intelligence agencies at times
- L2: Collected as they appear; indicators for a new vulnerability trickle in days to weeks after it is announced
- L3: Extracted from text by NLP and, lately, large language models, which are inconsistent and overconfident on real reports
way in: Sightings and false-positive reports from consumers
- L1: Consumers' detection systems and honeypots report when an indicator is seen
- L2: Sightings recorded, a few as confirmed false positives
- L3: Each new sighting resets the indicator's score; without one it decays
way in: The adversary's changes (new infrastructure, new tooling)
- L1: Attackers change their infrastructure and indicators over time
- L2: Mature indicators stop matching: the adversary must avoid them to succeed
- L3: Each kind of indicator ages at its own rate
way in: The customers' questions (requirements, requests for information)
- L1: Leaders' questions about business risk: who targets the sector, geopolitics, trends, impact
- L2: Requests for information, budgeted in vendor contracts at about ten a year
way in: Trusted sharing communities (ISACs, peer groups, sharing platforms)
- L1: Members' own threat information, exchanged without payment in formal (ISAC) or informal communities of trust
- L2: Labelled by the originator with how far it may travel (Traffic Light Protocol)
- L3: Exchanged machine to machine (STIX over TAXII, sharing platforms such as MISP)
way out: Attribution statements (with confidence) for law enforcement and policy
- L1: Used in investigations and policy responses
- L5: Assessments in low, medium or high confidence; errors carry diplomatic, legal or strategic costs
way out: Courses of action for detection, mitigation and response (hunting and incident response)
- L1: Detection and courses of action moved up the kill chain, before the exploit
- L6: Each phase of an intrusion mapped to courses of action
way out: Public reports naming actors
- L1: Actors named in each vendor's own scheme
- L5: A report released with the actor's fingerprints: domains, IP addresses, file hashes (APT1, 2013)
way out: Sharing back to communities (ISACs, peers), TLP-labelled
- L1: Indicators shared in a community of trust
- L3: Encoded in STIX, the most used sharing standard
- L5: The originator labels each item with one of four colours
way out: Strategic briefings to leaders
- L1: Priorities set and security resources allocated
- L7: Reports, briefings and risk assessments for executives and boards
way out: Tactical indicators into blocking and detection tools
- L1: Matching activity on the network flagged or blocked
- L3: Pushed through integrations into SIEM, EDR and response platforms
- L5: Tactical product: IP addresses, domains, hashes for SIEM rules, firewall policies and endpoint tools