Many security teams run threat intelligence and threat hunting as two separate functions that rarely talk, so intelligence reports go unread by hunters and hunt findings never reach the people who set intelligence priorities. Threat intelligence works outward: it studies adversaries, their intent and their behaviour, and turns that into assessments and requirements. Threat hunting works inward: it searches your own telemetry for adversary activity that existing detections missed. This guide compares the two on purpose, inputs, outputs, owners and failure modes, and spends most of its length on the hand-off loop between them: how intelligence requirements seed hunt hypotheses and how hunt findings flow back as intelligence. It is for CTI leads, hunt team leads and SOC managers. It does not re-explain either discipline from scratch: for the foundations, see the Threat Intelligence Guide and the Threat Hunting Guide.
Key takeaways
- Intelligence asks "who and how", hunting asks "are they here". One produces knowledge about adversaries; the other tests that knowledge against your own environment.
- The value is in the loop, not either half. Intelligence without hunting stays theoretical; hunting without intelligence drifts towards whatever the hunter happens to find interesting.
- An indicator sweep is not a hunt. Searching logs for a list of hashes and IP addresses is retrospective matching, useful but shallow; a hunt tests a behavioural hypothesis that does not depend on indicators the adversary can change cheaply.
- Hunt findings are first-party intelligence. What adversaries actually did in your network, which telemetry was missing and which techniques your detections miss is some of the most relevant intelligence you will ever hold.
- Measure the hand-offs. Track how many hunts trace to an intelligence requirement and how many hunt outputs became detections, intelligence updates or risk items, not just how many hunts ran.
Short answer
Threat intelligence and threat hunting are not alternatives. Intelligence is the knowledge function: it collects, evaluates and analyses information about threats so that people and systems can make better decisions. Hunting is a detection function: analysts proactively search for adversary activity on the assumption that some has already evaded automated controls. NIST SP 800-53 Rev. 5 treats them as separate controls, PM-16 Threat Awareness Program for intelligence sharing and RA-10 Threat Hunting for the hunt capability, and NIST CSF 2.0 places intelligence intake in ID.RA-02 and its use in analysis in DE.AE-07. If you can fund only one, a small organisation usually gets more from a modest intelligence capability that drives better detections and a quarterly hunt from a service provider than from an in-house hunt team with no intelligence input. A mature programme runs both with a formal, measured loop between them.
Side-by-side comparison
| Dimension | Threat intelligence | Threat hunting |
|---|---|---|
| Core question | Who threatens us, how do they operate, what should we prioritise? | Is that activity, or something like it, present in our environment now? |
| Direction of focus | Outward: adversaries, campaigns, sector trends | Inward: your endpoints, identities, network and cloud telemetry |
| Primary inputs | Intelligence requirements, open and commercial sources, sharing communities, incident reports, hunt findings | Hypotheses, telemetry (EDR, identity provider, DNS, proxy, cloud audit logs), baselines, intelligence |
| Primary outputs | Assessments with stated confidence, adversary profiles, indicators, priority technique lists | Confirmed or refuted hypotheses, incidents, new detections, visibility gaps, hunt reports |
| Typical owner | CTI team, often within the security operations or risk function | Hunt team or senior SOC analysts, sometimes a managed service |
| Consumers | Executives, risk owners, SOC, hunters, vulnerability management | Incident response, detection engineering, CTI, platform owners |
| Cadence | Continuous collection, periodic finished reports, ad hoc alerts | Planned hunt cycles plus urgent hunts triggered by new intelligence |
| Success looks like | Decisions changed: controls funded, detections prioritised, risks re-rated | Adversaries found earlier, detections added, blind spots closed |
| 800-53 Rev. 5 anchor | PM-16 Threat Awareness Program; RA-3(3) Dynamic Threat Awareness | RA-10 Threat Hunting |
Threat intelligence in this pairing
Threat model: intelligence exists because defenders cannot cover every technique equally. Its job is to narrow attention to the adversaries and behaviours most likely to affect your organisation, given your sector, geography, technology and exposure.
Architecture in words: intelligence requirements, agreed with stakeholders, drive collection from feeds, sharing groups, vendor reporting and internal incidents. A threat intelligence platform stores indicators and links them to adversary profiles and reports. Analysts produce finished products for different audiences: strategic summaries for leadership, operational campaign assessments for the SOC and hunters, and tactical indicators for machine consumption. NIST SP 800-150 describes the two sides of this work as consuming threat information from sources that fit your goals and producing it for sharing communities under agreed rules.
What hunters need from it: not a list of indicators but a prioritised set of techniques, with context on how a given adversary uses them (the tools, the order, the target systems) and a confidence level. A statement such as "groups targeting our sector are abusing OAuth application consent and then using application access tokens for lateral movement (T1550.001)" seeds a hunt. A list of 400 IP addresses seeds a sweep.
Failure modes: collection without requirements; reports written for the analyst's interest rather than a decision; indicators pushed to the SIEM without expiry; no channel to learn what hunters and responders actually saw. See Threat Feeds vs Finished Intelligence for the feed versus analysis trade-off.
Threat hunting in this pairing
Threat model: hunting assumes compromise. Preventive and detective controls have gaps, and an adversary using valid accounts (T1078) or built-in tools may produce no alert at all. RA-10 frames the capability as detecting, tracking and disrupting threats that evade existing controls.
Architecture in words: hunters need searchable, retained telemetry (endpoint process and command line data, identity provider sign-in and audit logs, DNS and proxy logs, cloud control plane logs), a place to analyse it at scale (SIEM, data lake or notebooks), and a documented method. The PEAK framework from SURGe by Splunk names three hunt types: hypothesis-driven, baseline (exploratory analysis of normal behaviour) and model-assisted. Its phases are Prepare, Execute and Act, and it treats threat intelligence as one form of the "knowledge" that feeds every phase, alongside business context, hunter experience and findings from the current hunt.
What intelligence needs from it: evidence. Which techniques were observed, on which systems, with what tooling, at what stage, and which techniques were searched for and not found given adequate visibility. Equally important: where visibility was too poor to say either way.
Failure modes: hunts chosen by habit rather than priority; no record of what was searched and why; findings closed as "nothing found" with no statement of data coverage; no conversion of a successful hunt into an automated detection. See the Threat Hunting Frameworks Guide for methods in depth.
The hand-off loop: intelligence to hunt
The forward path turns intelligence into testable questions. It works best as a defined workflow rather than an informal conversation.
- Start from a requirement. A priority intelligence requirement such as "which intrusion sets target our sector's cloud identity estate, and how?" gives the CTI team a reason to collect and gives the hunt team a reason to listen.
- Translate the assessment into techniques. CTI maps the adversary behaviour to ATT&CK technique IDs, with the procedure detail that matters: which binaries, which accounts, which log sources show it. MITRE's TTP-Based Hunting paper argues that collecting and filtering data around adversary techniques is effective because the techniques available on a target operating system are limited, so adversaries must reuse known ones or invest heavily in new ones.
- Check visibility first. Before a hunt is scheduled, the hunt team confirms the required telemetry exists and is retained long enough. If it does not, that gap is itself an output, routed to the platform owner and recorded as a risk.
- Write the hypothesis. A good hypothesis names the behaviour, the scope and the evidence that would confirm or refute it: "If an adversary is harvesting credentials from LSASS memory (T1003.001) on our server estate, we will see non-standard processes opening handles to lsass.exe with memory read access in EDR telemetry over the past 90 days." Record it in the Threat Hunt Hypothesis Backlog Template.
- Prioritise the backlog. Rank hypotheses by intelligence confidence, relevance to your crown jewels, current detection coverage and data readiness. Urgent intelligence (an active campaign against peers) can jump the queue.
- Agree the escalation path. Hunters must know when to stop hunting and open an incident. See Threat Hunting vs Incident Response.
The hand-off loop: hunt back to intelligence
The return path is the half most programmes neglect. Hunt teams produce first-party evidence about adversaries and about your own defences, and CTI is the function best placed to turn it into reusable knowledge.
| Hunt output | Intelligence use | Other destination |
|---|---|---|
| Confirmed malicious activity | New or updated adversary profile, internal indicators and procedures, sharing with peers under TLP | Incident response case |
| Technique observed, procedure differs from public reporting | Correct the technique profile; raise confidence in the actor assessment or record a new variant | Detection engineering backlog |
| Hypothesis refuted with good visibility | Lower the near-term likelihood estimate for that technique in this environment, with a review date | Coverage record |
| Hypothesis untestable through lack of data | Flag that intelligence on this technique cannot currently be acted on | Logging gap and risk register entry |
| Benign but unusual behaviour | Baseline context for future analysis | Allow-listing and documentation |
| Misconfiguration or exposure | Input to exposure and targeting assessments | Vulnerability or posture management |
Write the return path down. A short structured field set in every hunt report (techniques tested, data sources and time range, result, confidence, intelligence requirement served) lets CTI aggregate across hunts. The Threat Hunt Hypothesis and Report Template provides a starting format.
Why an IOC sweep is not a hunt
The most common confusion between the two disciplines is calling an indicator search a hunt. When a vendor publishes a report with hashes, domains and IP addresses, someone searches the SIEM for them and reports "hunt complete, nothing found". That activity has value as retrospective matching, but it is limited in three ways.
- Indicators are cheap to change. The Pyramid of Pain shows hashes, IP addresses and domains at the bottom because adversaries change them with little effort. A clean sweep says only that this adversary did not reuse these exact artefacts here. See the Pyramid of Pain Guide.
- "Nothing found" is not a finding without coverage. If DNS logs are retained for 14 days and the campaign started 60 days ago, the sweep could not have found it.
- It skips the analysis. A sweep does not ask how the adversary operates or whether a variant of that behaviour exists. A hunt built from the same report would extract the techniques and search for the behaviour, whatever the indicators.
Keep both, but name them correctly. Indicator matching belongs to automated detection (feeds into the SIEM, EDR and DNS controls with expiry) and quick retrospective searches. Hunting is reserved for hypothesis-driven work. The distinction matters for reporting: counting sweeps as hunts inflates hunt metrics and hides the fact that no behavioural hunting is being done. Note that RA-10 itself includes searching for indicators of compromise as one part of the capability, alongside detecting threats that evade existing controls; an RA-10 programme that only does the first part meets the letter but not the intent. For the indicator versus behaviour distinction, see Indicators of Compromise vs Indicators of Attack.
Choosing where to invest: decision criteria
| Situation | Lean towards | Why |
|---|---|---|
| Small team, limited telemetry retention | Intelligence-driven detection, external hunts | Hunting needs data and time; a few priority techniques turned into detections gives more coverage |
| Mature SOC with good EDR and identity logs | In-house hunting fed by CTI | The data exists; the gap is finding what rules miss |
| Frequently targeted sector, active peer campaigns | Both, with an urgent-hunt path | Intelligence on peers gives near-term hypotheses worth testing quickly |
| Leadership asks "are we a target?" | Strategic intelligence first | Hunting cannot answer intent or targeting questions |
| Leadership asks "have we been hit?" | Hunt or compromise assessment | Only evidence from your environment answers it; see the Compromise Assessment Guide |
| Detections noisy, coverage unknown | Detection engineering before more hunting | See Threat Hunting vs Detection Engineering |
Trade-offs: intelligence scales across the whole programme (one good assessment can reprioritise vulnerability management, detection and awareness) but produces no direct evidence about your environment. Hunting produces evidence but is labour-intensive and only covers the hypotheses tested. Large organisations can staff both and rotate analysts between them, which is the simplest way to keep the loop alive. Smaller ones often place a single analyst across both roles, which works if the hunt backlog is explicitly fed from intelligence requirements.
Worked scenario: one report, two responses
A government advisory describes an intrusion set targeting your sector. It lists indicators and describes behaviour: initial access through phishing (T1566), PowerShell for execution (T1059.001), lateral movement over RDP (T1021.001) using valid accounts, and persistence by adding credentials to cloud applications (T1098).
Response A, sweep only: the SOC loads the indicators into the SIEM, searches 30 days of logs, finds nothing and closes the ticket. Leadership is told the organisation is not affected.
Response B, the loop: CTI assesses relevance (sector match, same cloud platform, moderate confidence) and passes three priority techniques to the hunt team with procedure detail. The hunt team confirms 90 days of EDR and identity provider audit data, then runs three hypotheses: unusual RDP between servers by accounts that normally do not use it; encoded PowerShell launched from office applications; new credentials or secrets added to application registrations outside change windows. The first two are refuted with good coverage. The third finds two unexplained client secrets added to an application with broad mail permissions. Incident response takes over; the credentials are revoked. CTI updates its assessment (the technique is in active use against your estate), shares sanitised procedure detail with the sector group under the agreed TLP level, and detection engineering builds a rule for credential additions to privileged applications. The risk owner records the gap in application change monitoring.
The same report produced one ticket in Response A and five durable improvements in Response B.
Where identity fits
Much current adversary tradecraft is identity tradecraft: valid accounts, token theft, OAuth consent abuse and credentials added to service principals. Intelligence on these techniques is only actionable if hunters can see identity provider audit logs and non-human identity activity, which are often outside the traditional SOC data set. Make identity telemetry a standing requirement in both functions. See the Identity Threat Hunting Guide and Threat Intelligence for Identity and NHI Guide.
Implementation roadmap
- Agree shared requirements (month 1). CTI and hunt leads agree three to five priority intelligence requirements and a shared technique priority list.
- Define the hand-off artefacts (month 1). A hunt request format from CTI and a structured findings block in every hunt report.
- Run a pilot cycle (months 2 to 3). Three intelligence-driven hunts end to end, including a visibility check before each and a retrospective after.
- Wire the outputs (month 3). Route findings to IR, detection engineering, the risk register and the intelligence platform; agree owners for each route.
- Establish cadence (month 4 onwards). A fortnightly or monthly joint review of the backlog, plus an urgent-hunt trigger for high-confidence intelligence on active campaigns.
- Review and adjust (quarterly). If hunts are not producing outputs, check whether the intelligence was specific enough or the data was adequate before changing either team. Roll back to a simpler cadence rather than abandoning the loop if workload becomes unsustainable.
Detection and measurement
- Intelligence-sourced hunts: percentage of hunts that trace to a documented intelligence requirement or report.
- Hunt-to-detection conversion: percentage of completed hunts that produced at least one new or improved automated detection.
- Return-path rate: percentage of hunt reports whose findings were ingested by CTI (profile updated, confidence changed or shared).
- Visibility gaps closed: number of telemetry gaps identified by hunts and closed within an agreed period.
- Time from intelligence to hunt: median days from a high-priority intelligence product to a completed hunt on it.
- Behavioural share: ratio of hypothesis-driven hunts to indicator sweeps, reported separately.
See the Threat Intelligence Programme Metrics Guide for wider CTI measures.
Common failure patterns
- Sweeps reported as hunts: hunt metrics look healthy while no behavioural hunting happens.
- One-way traffic: CTI publishes, hunters consume, nothing comes back, so the intelligence picture never reflects your own environment.
- "Nothing found" without coverage: refuted hypotheses recorded without data sources and time range, giving false assurance.
- Intelligence too generic to hunt: reports that name actors but not procedures, leaving hunters to research from scratch.
- Successful hunts not automated: the same hypothesis is hunted manually every quarter instead of becoming a detection.
- Identity data missing: intelligence on token theft and consent abuse that hunters cannot test because identity logs are not collected.
Framework mapping
| Framework | Threat intelligence | Threat hunting |
|---|---|---|
| NIST CSF 2.0 | ID.RA-02 cyber threat intelligence is received; ID.RA-03 threats identified and recorded | DE.AE-02 adverse events analysed; DE.AE-07 threat intelligence integrated into analysis |
| NIST SP 800-53 Rev. 5 | PM-16 Threat Awareness Program; RA-3(3) Dynamic Threat Awareness | RA-10 Threat Hunting; SI-4 System Monitoring |
| ISO/IEC 27001:2022 Annex A | A.5.7 Threat intelligence | A.8.16 Monitoring activities |
| MITRE ATT&CK | M1019 Threat Intelligence Program; technique profiles of relevant groups | Techniques as the unit of hypothesis and coverage |
Practitioner checklist
- Agree priority intelligence requirements that the hunt team helped write.
- Express intelligence for hunters as ATT&CK techniques with procedure detail and confidence.
- Confirm telemetry and retention before scheduling any hunt; record gaps as risks.
- Write every hunt as a falsifiable hypothesis in a shared backlog.
- Report indicator sweeps separately from hypothesis-driven hunts.
- Add a structured findings block to every hunt report and route it to CTI.
- Convert repeatable successful hunts into automated detections.
- Include identity provider and non-human identity telemetry in hunt scope.
- Review the loop metrics quarterly with both team leads.
Standards and references
- NIST: SP 800-150 Guide to Cyber Threat Information Sharing
- NIST: SP 800-53 Rev. 5 Security and Privacy Controls (PM-16, RA-10)
- NIST: Cybersecurity Framework 2.0
- MITRE: TTP-Based Hunting
- MITRE ATT&CK: M1019 Threat Intelligence Program
- Splunk SURGe: The PEAK Threat Hunting Framework
- David Bianco: The Pyramid of Pain
Related NHI Mgmt Group resources: Threat Hunting Guide, Threat Intelligence Guide, Threat Feeds vs Finished Intelligence, Indicators of Compromise vs Indicators of Attack, Strategic vs Operational vs Tactical Threat Intelligence.