1. LandOffer
  2. Blog
  3. All articles
  4. Threat Intelligence Portfolio: Turn Public Reporting into an Evidence-Based Assessment

Job search

Threat Intelligence Portfolio: Turn Public Reporting into an Evidence-Based Assessment

Build a claim-level public-report ledger, explain source dependencies and uncertainty, and write an assessment with alternatives and decision indicators.

10 min readLandOffer team

A threat intelligence portfolio should answer a defined reader question with traceable reporting, an explicit judgment and the evidence that could change it. Collecting malware names or copying indicators does not show how you assess a source. Give a reviewer enough information to distinguish what a publisher reported, what you inferred and what remains unknown.

LandOffer publishes this guide; use recent tech job listings to identify the intelligence responsibilities your sample should demonstrate. This project reads two real 2026 vendor research reports and builds an original evidence ledger, then a separate assessment for a fictional service-desk decision. Sources were inspected October 23, 2026. No malware, suspicious domain, private telemetry, customer account or threat-actor infrastructure was accessed. The work demonstrates public-source analysis, not a confirmed incident or a production intelligence program.

An analyst compares two publisher folios, a shared reference thread and separate evidence and judgment trays.

Define a question that the available reporting can answer

Start with the consumer, decision and information cutoff. “Write about macOS malware” invites a summary with no clear purpose. “Does the selected reporting justify reviewing our assumption that a clean automated page result establishes download safety?” creates a testable analytical question. That question tests the sufficiency of an observation; it does not establish exposure at a particular organization.

For the first asset, the reader is a portfolio reviewer evaluating source handling. The question is whether two publications provide independently corroborated evidence of one campaign. For the second, the fictional reader manages a service desk deciding whether to revise a software-download review procedure. Those are separate decisions; neither requires inventing an attack against the reader’s organization.

The ODNI ICD 203 document describes source quality, uncertainty, distinctions between information and judgments, alternatives and customer relevance. Its stated applicability concerns the Intelligence Community. Here those ideas inform an editorial method; this portfolio is not certified by ODNI or claimed to satisfy an agency’s evaluation program.

Define the sample’s limits before collecting more material. This one evaluates the two reports as available on the inspection date. It does not claim an exhaustive threat landscape, a malware prevalence estimate, current infrastructure status or access to either vendor’s underlying telemetry. Publication date and the period of reported observations belong in different fields.

Worked asset one: a completed public-report provenance ledger

The ledger contains two primary research documents, not two invented attacks. S1 is Sophos’s Evil evolution: ClickFix and macOS infostealers. S2 is Microsoft’s August 5, 2026 report on a macOS ClickFix campaign’s cloaking gate. Their substantive report bodies were read, including qualifications and references. Reading a report does not reproduce its experiments or validate its telemetry.

Record Publisher-reported information Provenance and scope Assessment boundary
S1-A Sophos describes three campaigns across November 2025, December 2025 and February 2026 First-party research narrative; historical observations within a report identified as March 11, 2026 in S2’s references Three described campaigns are not three independently verified incidents in this portfolio
S1-B A reported counter measured copying an installation command, not confirmed infection S1, December campaign discussion and table qualification Do not relabel its click counts as compromised devices or add snapshots together
S1-C Sophos says it does not know whether the three campaigns had the same operator S1, introduction to the campaign comparison Shared technique or malware label does not establish one actor
S2-A Microsoft describes a shift from open lures to environment-dependent content delivery S2, activity overview and delivery-phase discussion This reports a specific tracked operation, not every suspicious download page
S2-B Microsoft describes benign or empty responses for some analysis environments S2, delivery and gate discussion A clean-looking response alone cannot establish safety in that reported setting
P-01 S2 explicitly lists S1 among its references S2, References Two publisher names do not establish independent underlying datasets

The completed source-quality note is: “Both are identifiable first-party research publications with described observations and technical reasoning. Their underlying collection and complete denominators are unavailable here. S2’s reference to S1 is visible; the extent of shared evidence is unresolved. Treat them as two documents with a documented relationship, not two proven independent confirmations of one operation.”

The ledger’s conclusion is deliberately narrower than a campaign attribution: the source set supports examining social-engineering delivery and visibility limits. It does not establish common ownership of the campaigns, a shared victim population or the probability that a particular company was compromised. A later article repeating S1’s account would add a publication, not automatically another observation.

Attach each citation to the particular claim it supports so a reviewer can check the relationship. “According to Sophos” is insufficient if a paragraph combines a stated observation with the portfolio author’s explanation. Keep S1-B attached to the counter’s meaning, and label any proposal about an organization’s review process as your own judgment. The table is a completed analytical product, not a collection of empty fields.

Separate source credibility, judgment confidence and event likelihood

Source credibility asks what access, expertise, method and possible limitations a publication provides. Confidence describes how strongly the available evidence and reasoning support your particular judgment. Likelihood concerns an event or development occurring. They answer different questions; none can be substituted for another merely because the publisher is well known.

For this project, confidence is high that S2 publicly describes conditional responses: that claim can be checked in the saved publication. Confidence is limited about how completely these two reports describe the wider macOS threat landscape. The source set supplies no defensible numerical probability of compromise for the fictional service desk. “High confidence” in reading a paragraph is not “high likelihood” of an intrusion.

Explain each label beside its subject. Avoid a single badge saying “source confidence: high” across the whole table, which could appear to validate every technical claim and inference. Describe the reason: identifiable publisher, explicit observation, visible qualification, inaccessible underlying data, dated collection or a documented reference relationship. A label without that explanation hides the reasoning a portfolio is supposed to expose.

Use uncertainty to identify a decision boundary. An unknown denominator prevents estimating a population infection rate. An unknown relationship between source datasets prevents multiplying corroboration. Neither gap prevents stating what the published text says. The reader should be able to see which conclusions survive the missing information and which ones do not.

The project includes an original local checker, the completed ledger and a separate assessment record. It checks that each reported claim names a known source, includes a location and finds its short reference anchor in the saved report receipt. It also checks that inferences stay labeled, the reference relationship is retained and neither an actor nor a compromise probability has been assigned.

Two fresh local executions produce matching output files. Negative tests substitute an unknown source ID, point to an absent anchor and relabel the documented citation as independent confirmation. Each altered record is rejected. This shows what the small checker actually validates: traceability and declared analytical constraints in these authored inputs.

It does not verify a vendor’s detections, establish dataset independence or authenticate an incident. A matching text anchor can support a claim-location check while the interpretation still needs human review. Saved file hashes identify which report receipts and authored artifacts were reviewed; they are not certificates that a published claim is true.

Keep this distinction in the README and portfolio summary. Do not say “validated ClickFix infrastructure” when the script only checked source IDs. Do not turn a deliberately rejected malformed record into evidence that a security product detects malware. No live lookups, payload retrieval, command execution or network scans are necessary for this exercise.

Worked asset two: a short assessment for a different decision

The following is the completed ASSESS-02, prepared October 23, 2026 for the fictional Cedar Quay service-desk manager. Its decision is whether to require additional review before treating an automated page result as sufficient evidence to approve a software-download route. Cedar Quay has supplied no actual URL result, endpoint event or user report; the scenario is a prospective procedure review, not a simulated incident masquerading as one.

Key judgment: A clean-looking automated response should not, on its own, be treated as proof that a proposed software-download route is safe. This is a recommendation about evidence sufficiency, not a claim that a particular download is malicious. The scenario assumes an approved-software route exists. The hypothetical manager should retain it while asking the authorized security owner what additional validation the procedure requires.

Basis: S2’s account supplies a concrete reason that one observation context may not reveal the content delivered in another. S1 supplies a different measurement caution: a recorded interaction is not necessarily installation or compromise. These are different limitations, not a combined infection-rate calculation. S2 cites S1, and shared source data have not been ruled out.

Confidence and likelihood: Confidence is moderate in the proposed procedure change’s relevance to Cedar Quay because the reasoning is supported by published cases, but the organization’s tooling and workflow are unknown. No likelihood of Cedar Quay compromise is estimated. The assessment does not assign a threat actor, malware family or malicious status to an unprovided URL.

Alternatives and what would distinguish them:

Possible explanation for an apparently benign response Evidence needed in an authorized review Consequence for this assessment
The destination is an authentic, approved software source Confirmed vendor provenance and an approved distribution route May resolve the download question without a campaign conclusion
The check was incomplete or the page was temporarily unavailable Tool coverage, time, redirect handling and the exact result meaning Keep the result inconclusive; do not label absence of findings as safety
Content differs by visitor context, as reported in S2’s case Authorized security analysis establishing context-dependent content for the actual destination Would strengthen the need for a specialized review; not established here

Indicators that would change the judgment: An authorized security owner documenting that the proposed review already accounts for relevant delivery contexts could reduce the need for a new step. Evidence that a requested installer came through an unapproved route would support reviewing that route under existing policy. Actual execution or credential-access evidence would move the matter into incident handling; none was collected for this assessment.

Actions and ownership: The service-desk manager may propose the procedure question; the authorized security owner determines the validation approach and any operational response. No block, account reset, software installation or user contact has occurred. Do not visit suspicious pages to make the portfolio more convincing. Preserve the proposed action and the evidence needed to authorize it as separate records.

Limitations and review point: The assessment uses two dated public research narratives, not Cedar Quay telemetry, a measured false-negative rate or current malicious infrastructure verification. Revisit it when the actual procedure, tool coverage or additional relevant reporting is available. No fixed review interval is presented as a universal requirement.

A service-desk assessment notebook weighs three alternative explanations, with empty evidence slots and a separate proposed-action card.

Package the analysis so someone else can challenge it

Give the reviewer one project index linking the question, cutoff date, source register, claim ledger, assessment and checker. Provide the precise report URLs and reviewed receipt hashes. Mark direct-fetch failures accurately: in this run, Microsoft’s direct request returned 403 and Sophos’s timed out, while their substantive bodies were read through the web research tool. An unrelated bulletin’s script-only shell was not promoted into supporting report evidence.

For a reviewer, the useful demonstration is not the number of sources collected. It is the ability to follow a claim, see a dependency, locate an uncertainty and understand a decision that could change. Include the rejected records and the limits of the checks. Remove private identifiers and do not copy customer logs into a public portfolio.

A truthful summary is: “Built a public-report evidence ledger and a bounded service-desk assessment; preserved source dependencies, competing explanations and decision indicators, with local traceability checks.” It does not claim threat hunting, victim discovery, malware reverse engineering or containment. For a role in a LandOffer job shortlist, identify which requirement this sample addresses and show your authored reasoning alongside the publishers’ separately attributed observations.

Sources and Further Reading