Threat hunters are hired to find the intrusions that every alert missed, which makes it one of the hardest security jobs to get into and one of the easiest to fake on a CV. Hunting is proactive, hypothesis-led searching of telemetry for adversary behaviour, and it rewards people who know operating systems, identity and networks well enough to spot the unusual in the normal. This guide covers what hunters actually do, the skills and tools interviewers test, the routes most hunters take (almost always through a SOC, incident response or detection engineering), which certifications help and where the role leads. It is written for SOC analysts and responders planning a move into hunting, and for managers building a hunt function. Hunting methods are covered in the Threat Hunting Frameworks Guide; broader career routes are in the Cybersecurity Careers Guide.
Key takeaways
- Hunting is a second or third job, not a first one. Almost every hunter arrives from SOC tier 2 or 3, incident response, forensics or detection engineering, because the job assumes you already know what normal looks like.
- Behaviour beats indicators. MITRE's TTP-based hunting work argues that adversaries are constrained by the technology they attack, so hunting for techniques is more durable than hunting for hashes and IP addresses.
- The output is detections and fixes, not just findings. A good hunt ends with a new or tuned rule, a closed visibility gap or a hardening change, written up so others can repeat it.
- Query fluency is the daily tool. KQL, SPL or another SIEM language, plus Python for the data that does not fit, is what hunters use every hour; interviews test it live.
- Identity is the busiest hunting ground. Valid account abuse, token theft and service account misuse leave little endpoint trace, so hunters who read IdP and cloud audit logs are in demand.
What the role does
NIST SP 800-53 Rev. 5 has a dedicated control for the activity, RA-10 Threat Hunting, alongside SI-4 System Monitoring and AU-6 Audit Record Review, Analysis, and Reporting. The NICE Framework (components v2.2.0, April 2025) has no separate hunter work role; the work spans PD-WRL-001 Defensive Cybersecurity (task statements include "Detect cybersecurity attacks and intrusions" and "Analyze network traffic anomalies") and PD-WRL-006 Threat Analysis. Job titles include threat hunter, cyber threat hunter, hunt team lead and, in many SOCs, tier 3 analyst with a hunting remit.
MITRE's 2020 paper TTP-Based Hunting gives the reasoning behind modern practice: the number and type of techniques adversaries can use after compromise is limited by the target's technology, so "All adversaries must either employ these known techniques or expend vast resources to develop novel techniques". Hunters therefore organise work around ATT&CK techniques and the data sources that reveal them.
| Hunt type | Starting point | Example | Typical output |
|---|---|---|---|
| Hypothesis-driven | A testable statement about adversary behaviour | "An attacker with a stolen credential is using RDP (T1021.001) from a workstation that has never used it" | Finding or confirmed absence, new detection, logging gap list |
| Intelligence-driven | A report on a group or campaign relevant to your sector | Search for the scheduled task (T1053.005) and run key (T1547.001) persistence patterns described in an advisory | Scoped answer to "are we affected?", indicators added to blocking |
| Baseline or anomaly | Statistical view of normal | Rare parent and child process pairs, rare service account logon hosts | Baseline documentation, outliers triaged, noisy software allow-listed |
| Crown-jewel focused | The systems that matter most | Credential access against domain controllers, such as LSASS memory access (T1003.001) | Hardening recommendations, high-fidelity detections |
A hunt from start to finish
- Pick and scope: choose the technique, environment and time window; write the hypothesis and what evidence would confirm or refute it. Use the Threat Hunt Hypothesis and Report Template.
- Check visibility: confirm the data source exists and is complete, for example Windows 4688 with command line, Sysmon event 1 or EDR process telemetry; missing data is itself a finding.
- Query and pivot: start broad, aggregate, then pivot on user, host, parent process and network destination.
- Triage outliers: explain each one as benign with a reason or escalate it to incident response under the normal process.
- Convert: hand durable logic to detection engineering as a rule with test cases. See the Detection Engineering Guide.
- Write up and share: hypothesis, data, queries, results, gaps and recommendations, so the hunt can be repeated next quarter.
Core skills and tools
| Skill | What good looks like | How to practise |
|---|---|---|
| Windows internals | Process trees, services, scheduled tasks, registry persistence, logon types and Kerberos ticket events (4768, 4769), LSASS protections | Generate each behaviour in a lab and find it in telemetry. See the Sysmon Guide |
| Linux and cloud | Auth logs, cron, systemd units, container runtime events, cloud audit logs | Run a small Linux host and a free cloud tenant with audit logging on |
| Identity telemetry | IdP sign-in and audit logs, Conditional Access results, OAuth consent, service principal and API key use, pass the hash (T1550.002) patterns | See the Identity Threat Hunting Guide |
| Query languages | Joins, time windows, stacking (least frequent occurrence), percentiles and baselines from memory | KQL for Threat Hunting Guide, Splunk SPL for Threat Hunting Guide |
| Data analysis | Python with pandas or notebooks for data that outgrows the SIEM; basic statistics for outliers | Rebuild a SIEM hunt in a notebook from exported logs |
| Network evidence | DNS, proxy, TLS metadata and flow logs; spotting beaconing and unusual destinations | Capture lab traffic and explain every conversation |
| ATT&CK and intelligence | Translate a report into techniques, data sources and testable hypotheses | Map three public advisories to hunts. See the Threat Intelligence Guide |
| Writing | Hunt reports a SOC manager can act on and a detection engineer can implement | Publish lab hunt write-ups |
Tools: SIEM and data lake query interfaces, EDR live query and process telemetry, Sysmon, network sensors, notebooks, Sigma for portable rule logic, and ATT&CK Navigator for coverage. Adversary emulation libraries, run only in a lab or with written authorisation, let you check that a hunt would have found the behaviour. Tools change; the skills above transfer.
Where identity fits
Valid Accounts (T1078) is the technique hunters most often chase, because an attacker who signs in with a real credential raises few endpoint alerts. Service accounts, workload identities and API keys are attractive because they rarely have MFA and their behaviour is regular, which also makes deviations easy to find once baselined. Hunting non-human identities is covered in the Detection Engineering for Non-Human Identities Guide.
Routes into the role
- SOC tier 2 or 3: the most common route. Volunteer for hunts, turn repeated investigations into hypotheses, and track the detections you contributed. See the SOC Analyst Career Guide.
- Incident response and DFIR: responders already think in attacker timelines; hunting is IR without a confirmed incident.
- Detection engineering: strong query and testing skills; add the exploratory mindset. See the Detection Engineer Career Guide.
- Threat intelligence: analysts who can query telemetry move into intelligence-driven hunting. See the Threat Intelligence Analyst Career Guide.
- System or network administration: deep knowledge of normal is valuable, usually via a SOC role first.
Market context: DSIT's Cyber security skills in the UK labour market 2026 found that only 16% of core cyber postings in 2025 were open to people with under a year's experience and 64% asked for two to six years; hunting roles sit at the experienced end. Incident response was requested in 18% of postings and automation in 15%. DSIT gives no hunter-specific salary; the mean advertised salary across core cyber roles was £60,800 in 2025. In the US, the Bureau of Labor Statistics reports a median of $129,180 (May 2025) for the broad information security analyst occupation, projected to grow 21% from 2025 to 2035.
Portfolio: three to five published hunt write-ups from your own lab (hypothesis, data, queries, results, detection produced) are the strongest evidence you can bring. Keep the lab isolated and only test systems you own.
Certifications that help
Certifications structure study and get past screening; interviewers still test live querying and reasoning. Check each body's current outline and price before booking. The table is descriptive, not a ranking.
| Awarding body | Certification | Relevance to hunting | Format facts (checked 27 October 2026) |
|---|---|---|---|
| CompTIA | CySA+ (CS0-004, V4) | Security Operations domain includes threat intelligence and threat-hunting concepts | Up to 85 questions, 165 minutes; domains Security Operations 34%, Vulnerability Management 26%, Incident Response and Management 24%, Reporting and Communication 16%; about four years' SOC or vulnerability analyst experience recommended. The V3 English exam retires 22 December 2026 |
| GIAC | GIAC Certified Forensic Analyst (GCFA) | Covers threat hunting and APT intrusion incident response | 82 questions, 3 hours, 71% to pass, includes CyberLive hands-on lab questions |
| GIAC | GIAC Cyber Threat Intelligence (GCTI) | Intrusion profiling, pivoting and intelligence that feeds hunts | 82 questions, 3 hours, 71% to pass |
| GIAC | GIAC Linux Incident Responder (GLIR) | Aimed at Linux incident response and threat hunting teams | See GIAC for current format |
| Microsoft or SIEM vendors | Platform security operations certifications | Query fluency in the tool your employer uses | Vendor specific |
Progression
| Next role | What changes | Build this while hunting |
|---|---|---|
| Hunt team lead | Owns the hunt backlog, prioritisation and reporting to the SOC manager | Hunt metrics, coaching, stakeholder briefings |
| Detection engineering lead | Owns rule quality and coverage across the estate | Rules as code, testing, ATT&CK coverage analysis |
| Incident response lead | Leads major incidents end to end | Scoping, containment authority, executive communication |
| Threat intelligence lead | Owns the intelligence requirements that drive hunts | Intelligence requirements, adversary tracking |
| Adversary emulation or purple team | Emulates techniques to test detection | Authorised testing discipline. See the Red Teamer Career Guide |
| SOC manager | People, budget, service levels | See the Security Operations Manager Career Guide |
Framework mapping
| Framework | Where hunting sits |
|---|---|
| NIST SP 800-53 Rev. 5 | RA-10 Threat Hunting, SI-4 System Monitoring, AU-6 Audit Record Review, Analysis, and Reporting |
| NIST CSF 2.0 | DE.CM Continuous Monitoring, DE.AE Adverse Event Analysis, ID.RA-03 threats identified and recorded |
| NICE Framework v2.2.0 | PD-WRL-001 Defensive Cybersecurity, PD-WRL-006 Threat Analysis |
| MITRE ATT&CK v20 | The common language for hypotheses, for example T1078, T1059.001, T1003.001 |
Measuring a hunter's impact
- Hunts completed and documented per quarter, with repeatable queries.
- Detections produced: rules that went to production from hunts, and their true positive rate.
- Visibility gaps closed: missing log sources found by hunts and later onboarded.
- Technique coverage: ATT&CK techniques hunted at least once in the last year on crown-jewel systems.
- Incidents found by hunting rather than alerts, and their dwell time.
Common failure patterns
- Hunting for indicators only: sweeping last week's hashes is useful hygiene, not hunting.
- No hypothesis: browsing data without a question produces nothing repeatable.
- Not checking visibility first: concluding "nothing found" from a data source that was never collected.
- Findings that die in a report: no rule, ticket or hardening change follows.
- Hunters pulled back into the queue: without protected time the hunt programme quietly stops.
Practitioner checklist
- Get to tier 2 or 3, IR or detection engineering first, and ask for hunt time.
- Become fluent in one SIEM query language and Python for log analysis.
- Learn Windows process, logon and Kerberos events and cloud sign-in logs in depth.
- Build an isolated lab with Sysmon and a SIEM, emulate techniques, and hunt for them.
- Publish three to five hunt write-ups using a consistent hypothesis and report format.
- Turn at least one hunt into a production detection and record its results.
- Pick one certification that matches your gap: query depth, forensics or intelligence.
- Include identity and service account behaviour in every hunt plan.
Standards and references
- MITRE: TTP-Based Hunting
- NIST: SP 800-53 Rev. 5, Security and Privacy Controls (RA-10 Threat Hunting)
- NIST: NICE Framework current versions
- MITRE ATT&CK: T1078 Valid Accounts
- CompTIA: CySA+ V4
- GIAC: Certified Forensic Analyst (GCFA)
- GIAC: Cyber Threat Intelligence (GCTI)
- DSIT: Cyber security skills in the UK labour market 2026
Related NHI Mgmt Group resources: Cybersecurity Careers Guide · Threat Hunting Guide · Threat Hunting Frameworks Guide · Identity Threat Hunting Guide