Stratir Field Guide

CTI operations checklist

Analyst workstation with a server room display

Designed by Stratir
The Bahamas & North America

Requirements, analysis, dissemination

CTI becomes valuable when it reduces uncertainty for a named decision, at the right moment, with evidence that can survive review.

Cyber threat intelligence is not an indicator feed.

It is a decision-support discipline. The value of a CTI function is not measured by how many reports it reads, how many hashes it stores, or how many technique IDs it maps. It is measured by whether the organization makes better security decisions because the intelligence function reduced uncertainty at the right time.

The first mistake in CTI operations is treating collection as the beginning of the workflow. Collection is not the beginning. The requirement is. A collection plan without a requirement becomes a stream of artifacts that someone later has to justify.

A good CTI operation begins by asking what problem needs to be solved. Is the team trying to detect an active intrusion, understand an extortion campaign, prioritize a vulnerability, protect an executive, investigate fraud infrastructure, support legal action, or brief leadership on risk? Each problem requires a different collection plan, different confidence threshold, and different product format.

Put every task inside the intelligence cycle.

Requirements, collection, processing, analysis, dissemination, and feedback are not interchangeable. They are different standards of work.

01

Requirement

Define the decision, stakeholder, risk scenario, time window, and intelligence gap before collection begins.

02

Collection

Collect from internal telemetry, incidents, open sources, trust communities, vendor reporting, malware analysis, and exposure data only when each source has a purpose.

03

Processing

Normalize observables, preserve raw evidence, deduplicate records, attach source metadata, and separate observed facts from analytic judgments.

04

Analysis

Convert processed data into judgments about adversary behavior, infrastructure, capability, targeting, opportunity, and likely impact.

05

Dissemination

Package intelligence for the audience, apply handling instructions, and distribute through a channel that matches the risk and sensitivity.

06

Feedback

Track whether detections fired, whether the intelligence changed a decision, and which requirements should be retired, rewritten, or escalated.

Requirements are the control plane.

A requirement should be written before collection begins, reviewed before analysis begins, and updated after feedback returns.

Requirement frame

Decision owner

Name the person or team that will act on the answer. If no one owns the decision, the request is not ready.

Requirement frame

Intelligence question

Write the requirement as a question that reduces uncertainty. Do not write it as a demand for every possible indicator.

Requirement frame

Risk scenario

Connect the question to an intrusion scenario, fraud pattern, exposure concern, geopolitical event, vendor risk, or executive decision.

Requirement frame

Time window

Define whether the product is tactical, operational, or strategic. A 30-minute incident need and a quarterly board need are not the same workflow.

Requirement frame

Source boundary

State which sources are authorized, which are excluded, and what legal, contractual, privacy, or OPSEC restrictions apply.

Requirement frame

Success criteria

Define what a useful answer looks like before collection begins. The team should know what would satisfy, revise, or close the requirement.

Evidence discipline

Preserve the record before shaping the story.

Processing is where weak CTI programs lose defensibility. A raw source record should not disappear after enrichment, summarization, or clustering. The original observation is the anchor that lets an analyst explain why a finding exists.

Treat indicators as evidence fragments, not conclusions. An IP address, domain, hash, wallet, selector, or malware family name is useful only when it is tied to source context, time, confidence, and the behavior it helps explain.

STIX and TAXII can help when structured objects need to move across platforms, but structure alone does not create intelligence. The structure has to carry provenance, relationships, sightings, confidence, and handling instructions.

Operational gate

Source gate

Every source has a reason to be collected, a reliability expectation, and a defined boundary. Volume is not the same as coverage.

Operational gate

Evidence gate

Raw records, timestamps, query terms, hashes, screenshots, enrichment outputs, and source URLs are preserved before analysis changes the shape of the data.

Operational gate

Processing gate

Selectors, entities, dates, infrastructure, malware names, vulnerabilities, and actor aliases are normalized before any conclusion is written.

Operational gate

Confidence gate

Every analytic claim has a confidence level, and low-confidence leads are labeled as leads rather than findings.

Operational gate

Action gate

The output tells detection, incident response, vulnerability management, leadership, legal, or fraud teams what decision it supports.

Operational gate

Feedback gate

The team records whether the intelligence was used, whether it was timely, and what collection gap still remains.

Analysis is where data becomes judgment.

The analyst is not just summarizing what sources said. The analyst is assessing what the evidence means, what remains uncertain, what action is justified, and what would change the assessment. This is the point where CTI must be careful with confidence, attribution, and implied urgency.

MITRE ATT&CK mapping should describe behavior, not decorate a report. A technique ID is useful when it clarifies what the adversary did, what telemetry should show, what detection logic can be written, and which control gap is exposed.

Attribution should be treated as a judgment, not a brand label. If the evidence supports malware family, infrastructure pattern, tooling overlap, language marker, targeting profile, or operational method, say that. If it does not support a named actor, do not force one.

Make analytic standards explicit.

A CTI product should read clearly, but it should also show how the analyst reached the assessment.

Analysis standard

Claim discipline

Each paragraph should make one claim, show the evidence behind it, and separate fact from assessment.

Analysis standard

Confidence language

Confidence should reflect source reliability, evidence quality, corroboration, recency, and analytic uncertainty.

Analysis standard

Competing hypotheses

A serious CTI product considers alternative explanations, especially when attribution, intent, or future action is being assessed.

Analysis standard

Behavior before labels

Actor names are less important than the behavior observed. Detection teams can act on behaviors even when attribution is unresolved.

Analysis standard

Impact translation

The assessment should translate technical activity into business risk, operational exposure, user harm, or executive decision context.

Analysis standard

Review record

The final product should preserve who reviewed it, what changed, what uncertainty remains, and when the requirement should be revisited.

03 / Dissemination

Send different products for different decisions.

Dissemination is not posting the same report to every channel. FIRST TLP exists because intelligence has handling constraints. A high-quality product still fails if it reaches the wrong audience, arrives without context, or lacks instructions for use.

A detection engineer needs behavior, telemetry, logic, and validation notes. A senior executive needs exposure, likelihood, business impact, confidence, and decision options. A legal or investigations team needs preserved evidence, source context, and review history. The format should follow the decision.

Deliverable

Executive risk note

A short decision memo for leadership that explains the threat, exposure, confidence, likely impact, and recommended action.

Deliverable

Detection package

A package of behaviors, indicators, logic, assumptions, false-positive notes, and validation criteria for detection engineering.

Deliverable

Hunt hypothesis

A testable statement about adversary behavior, where it should appear in telemetry, and what evidence would confirm or reject it.

Deliverable

Incident support brief

A time-sensitive brief that explains what is known, what is suspected, what is unknown, and which response decisions are waiting on intelligence.

Deliverable

Indicator feed

A structured feed for machine consumption only when indicators are validated, scoped, time-bound, and distributed with correct handling.

Deliverable

Strategic assessment

A longer assessment that connects campaigns, capabilities, targeting, geopolitical or criminal context, and organizational exposure.

Agentic CTI should accelerate evidence work, not replace judgment.

Agents can enrich, compare, cluster, draft, translate, and test. The analyst still owns confidence, attribution, risk, and release.

Agent role

Enrichment agent

Expands observables with passive DNS, certificates, malware reports, vulnerability context, abuse records, and internal telemetry references.

Agent role

Correlation agent

Clusters related records by shared infrastructure, selector reuse, temporal pattern, language, tooling, hosting, or victimology.

Agent role

Drafting agent

Turns structured notes into audience-specific drafts while preserving source references and uncertainty markers.

Agent role

Detection agent

Translates behaviors into hunt hypotheses, detection logic drafts, telemetry requirements, and validation notes for engineering review.

Agent role

Review agent

Checks claims against cited evidence, flags unsupported attribution, searches for missing context, and asks for confidence justification.

Agent role

Memory agent

Maintains requirements, prior findings, retired assumptions, source reliability notes, and feedback from previous intelligence cycles.

04 / Operating rhythm

A 72-hour rhythm keeps urgent CTI grounded.

01

0 to 4 hours

Frame the request

Confirm the decision owner, write the intelligence question, define sensitivity, and decide whether the request is tactical, operational, or strategic.

02

4 to 12 hours

Collect and preserve

Collect from scoped sources, preserve raw evidence, record collection paths, and avoid premature attribution while the evidence base is thin.

03

12 to 24 hours

Process and enrich

Normalize records, enrich observables, compare internal and external sources, and identify what is confirmed, probable, possible, or unresolved.

04

24 to 48 hours

Analyze and review

Build hypotheses, map behavior, challenge assumptions, record confidence, and run review before the product is released.

05

48 to 72 hours

Disseminate and measure

Send the right product to the right audience, track whether it changed a decision, and record what the next requirement should become.

Closing principle

A mature CTI program is not louder. It is more selective.

It collects less irrelevant data, preserves more context, produces fewer unsupported claims, and closes the feedback loop faster. The best checklist is not a box-ticking exercise. It is an operating discipline that makes every intelligence product easier to use, challenge, and improve.

Build with Stratir