Roadmaps Security roadmaps Defensive security and threat management

Threat Hunting

Search proactively for attackers that no alert has caught, starting from a hypothesis and ending with new detections.

Estimated time: 15–25 months · At about 10 hours a week.

Learning path

Click a topic to see what to learn and where. Mark topics as you go; your progress is saved in this browser only.

  • Core topic
  • Optional
  1. Beginner · 2–3 months Hunting basics
  2. Intermediate · 3–4 months Endpoint, network and identity hunts
  3. Advanced · 4–6 months Intelligence-driven and cloud hunting
  4. Expert · 6–12 months Leading a hunting programme

All topics in detail

BeginnerTime: 2–3 months

Hunting basics

Skills

  • Write testable hunting hypotheses from ATT&CK techniques.
  • Use stacking to find rare and suspicious values.
  • Document a hunt from start to finish.

Hands-on projects and labs

  • Load the Splunk Boss of the SOC dataset or a Security Datasets scenario and hunt for persistence by stacking.
  • Write and run three hunts based on ATT&CK techniques, and document each one.

Done when you can…

  • Find an attacker's persistence in a practice dataset without being told where it is.
  • Hand over a hunt report that someone else could repeat.

Beginner

What hunting is

Hunting is a human-led search for threats that automated detection missed. It is not the same as reviewing alerts or running indicator sweeps.

Resources

Beginner

Writing a hunting hypothesis

A good hypothesis is specific and testable: who, what technique, where in your environment, and which data would prove it.

Resources

Beginner

Data for hunting

Process starts, command lines, network connections, logons and DNS queries are the core. Know how long each is kept, since hunts often look back months.

Resources

Beginner

Stacking and frequency analysis

Count how often each value appears across many machines. Rare parent-child processes, services or autoruns are often where attackers hide.

Resources

Beginner

Hunting for persistence

Run keys, services, scheduled tasks, WMI subscriptions and startup folders. Compare them across hosts and against a known-good baseline.

Resources

  • Autoruns ToolFreeMicrosoft Sysinternals

Beginner

Documenting a hunt

Record the hypothesis, data, queries, findings and what changed. A hunt that found nothing is still useful if it is written down.

Resources

IntermediateTime: 3–4 months

Endpoint, network and identity hunts

Skills

  • Hunt across EDR, network and identity data.
  • Collect data from a whole fleet when the SIEM lacks it.
  • Find beaconing and data exfiltration in network logs.
  • Spot malicious use of built-in tools.

Hands-on projects and labs

  • Deploy Velociraptor to lab machines and hunt for a persistence technique on all of them.
  • Run RITA on Zeek logs from a lab where you simulated beaconing traffic.
  • Hunt for Kerberoasting in a lab Active Directory after running it yourself.

Done when you can…

  • Find a simulated intrusion in a lab using hunting alone.
  • Turn a successful hunt into a tested detection.

Intermediate

Hunting in EDR data

Use EDR telemetry to follow process trees, command lines and injected code across the whole fleet.

Resources

Intermediate

Fleet-wide collection

When the data you need is not in the SIEM, collect it from every machine at once, such as a list of scheduled tasks or a registry key.

Resources

Intermediate

Network hunting

Look for beaconing, long connections, rare destinations, odd DNS and data leaving at night. Zeek logs make this much easier than raw packets.

Resources

Intermediate

Identity and Active Directory hunting

Hunt for Kerberoasting, unusual logon types, new admins, password spraying and token theft in domain controller and identity provider logs.

Resources

Intermediate

Living off the land

Attackers use built-in tools such as PowerShell, WMI and certutil. Learn what normal use looks like so abnormal use stands out.

Resources

Intermediate

Notebooks for hunting

Jupyter notebooks let you query, join and visualise data and keep the whole hunt reproducible.

Resources

AdvancedTime: 4–6 months

Intelligence-driven and cloud hunting

Skills

  • Plan hunts from threat intelligence.
  • Hunt in cloud and SaaS audit logs.
  • Turn hunt results into lasting detections.

Hands-on projects and labs

  • Pick a recent DFIR Report case, extract the behaviours and hunt for each in a lab where you reproduced some of them.
  • Create a malicious mailbox rule in your own Microsoft 365 test tenant and hunt for it in the audit log.

Done when you can…

  • Run a hunt programme for a quarter with documented results and new detections.
  • Find something real that no existing alert had caught.

Advanced

Intelligence-driven hunts

Start from a report about a group that targets your sector, extract its behaviours and hunt for them in your data.

Resources

Advanced

Cloud and SaaS hunting

Hunt in CloudTrail, Azure and Microsoft 365 audit logs for new keys, mailbox rules, consent grants and unusual sign-ins.

Resources

Advanced

Hunting in memory

Some implants live only in memory. Use EDR memory scans or targeted memory captures to look for injected code.

Resources

Advanced

From hunt to detection

Every successful hunt should become a detection or an automated hunt that runs again. Hand it over with tests and documentation.

Resources

AdvancedOptional

Statistics and machine learning for hunting

Clustering and outlier detection can surface the unusual in very large data sets. Treat the results as leads for a human to check.

Resources

ExpertTime: 6–12 months

Leading a hunting programme

Skills

  • Run a hunting programme with clear goals and metrics.
  • Share hunts with the wider community.

Hands-on projects and labs

  • Write a hunting plan for a year, based on the threats to a chosen sector.
  • Publish a hunt write-up with queries others can run.

Done when you can…

  • Show that hunting lowered dwell time or improved detection coverage.
  • Have your hunts used by other teams.

Expert

Running a hunting programme

Decide what to hunt, how often, and how results feed detection and response. Hunting without a programme becomes random searching.

Resources

Expert

Measuring hunting

Count hunts run, incidents found, detections created and data gaps fixed. Dwell time across incidents shows whether it is working.

Resources

Expert

Sharing hunts

Publish hunt queries and write-ups so other teams can run them. Learn from theirs in return.

Resources

Overview

Threat hunting starts from the assumption that an attacker may already be inside and the alerts have missed them. A hunter forms a hypothesis, such as 'an attacker is using scheduled tasks for persistence', looks for evidence in endpoint, network and cloud data, and either finds something or proves the gap is covered.

Every hunt should leave something behind: a confirmed incident, a new detection, better logging, or a documented negative result. Hunting connects threat intelligence, detection engineering and incident response.

Why it matters

Skilled attackers avoid known detections. Hunting is how mature teams find them sooner, and how they keep improving their detections from real evidence.

Typical job roles

  • EntrySOC analyst (Tier 2) who hunts Runs guided hunts from playbooks and escalates findings.
  • Mid-levelThreat hunter Plans and runs hypothesis-driven hunts across endpoint, network and cloud data.
  • SeniorSenior threat hunter Leads hunts on advanced threats and turns results into detections.
  • LeadThreat hunting lead Runs the hunting programme, its method and its metrics.

Prerequisites

  • Security Operations intermediate stage.
  • Windows internals and common attacker techniques.
  • A query language such as KQL, SPL or SQL, and some Python.
  • Reading packet captures and network logs.

Cybersecurity Foundations of computer science

Tools and technologies

  • Velociraptor Open source Query and collect from every endpoint at once.
  • osquery Open source Query endpoints with SQL.
  • Zeek Open source Network logs built for analysis.
  • RITA Open source Finds beaconing and other suspicious network patterns in Zeek logs.
  • MSTICPy Open source Python tools for hunting in notebooks.
  • Jupyter Open source Notebooks for reproducible hunts.
  • Autoruns Free Shows every way a program can start on Windows.
  • Hayabusa Open source Fast timelines and hunting in Windows event logs.

Certifications

Take them in roughly this order. Costs and details for the ones marked with a link are on the Certifications page.

  1. Intermediate CySA+ (Cybersecurity Analyst) CompTIACovers threat hunting basics.Cost and exam details
  2. Intermediate Security Operations Analyst Associate (SC-200) MicrosoftHunting in Microsoft Sentinel and Defender.Cost and exam details
  3. Advanced GIAC Certified Forensic Analyst (GCFA) GIACAdvanced incident response and threat hunting.

Resources

Frameworks and standards

  • MITRE ATT&CK A knowledge base of real attacker tactics and techniques, used to plan detections, hunts and red team work.
  • MITRE D3FEND A knowledge graph of defensive techniques, mapped to the ATT&CK techniques they counter.
  • Cyber Kill Chain Seven stages of an intrusion, from reconnaissance to actions on objectives, used to plan where to detect and disrupt attacks.
  • NIST Cybersecurity Framework (CSF) 2.0 A common way to describe and improve an organisation's security, organised as Govern, Identify, Protect, Detect, Respond and Recover.

All frameworks and standards

Career path

A common progression. Titles vary between companies.

  1. SOC analyst0–2 years
  2. Tier 2 analyst or incident responder2–4 years
  3. Threat hunter4–7 years
  4. Threat hunting lead7+ years

Related domains to move into

All security roadmaps Back to top

Last reviewed: . Found a mistake, or want a roadmap written sooner? Contact