1. LandOffer
  2. Blog
  3. All articles
  4. Malware Analysis Portfolio: Present a Safe Report Without Distributing Live Samples

Job search

Malware Analysis Portfolio: Present a Safe Report Without Distributing Live Samples

A completed inert-text report and independent claim review show reproducible packaging while keeping static observations separate from untested behavior.

9 min readLandOffer team

A malware-analysis portfolio can give a reviewer a clear report, traceable observations and explicit unknowns without distributing a live sample. Identify what you examined, which methods you actually used and which conclusions the evidence supports. Keep executable specimens and unapproved case material out of the recruiting packet. A report should distinguish a behavior actually observed from one merely suggested by text.

For this guide, we created and parsed an inert plaintext fixture. It contains no malware, executable logic or real indicators of compromise. The completed report and claim-review artifacts demonstrate bounded reporting and reproducibility; they do not establish reverse-engineering, sandbox or live-malware experience. No specimen was downloaded, executed, uploaded or sent to a reviewer.

LandOffer publishes this guide. As you compare recent security roles, match their requested analysis work to the evidence you can explain. A careful reporting exercise belongs in a learning portfolio with its limitations visible, rather than being presented as a production malware investigation.

An inert paper specimen remains in a glass display case while a researcher reviews a separate report and evidence index. Editorial illustration.

Start with the report's question and evidence

Lenny Zeltser's report template, dated June 18, 2026, separates the executive summary, sample identity, sources, supporting analysis and what remains unknown. It also distinguishes the malware report from an incident-response or threat-intelligence report. That practitioner structure is useful here; it is not an employer's universal portfolio requirement or a certification of the work inside it. Zeltser's report structure

Give the reader a bounded question and result near the beginning. If your source is a public agency report, identify its exact title and publication date and distinguish the agency's observations from your own synthesis. Reading a published runtime finding does not mean you observed that behavior locally. If your input is synthetic, say so before the first finding.

For an authorized analysis you actually performed, name the methods and record their outputs. Do not fill an unused section with an inferred result simply because a template includes it. “Not performed” is more useful than a process-tree picture borrowed from another analysis and left unexplained.

A portfolio package should also explain your contribution. Did you collect permitted facts, interpret a source, write a parser, examine code or run an authorized experiment? Keep those verbs accurate. Reviewers can assess a small completed piece when its scope is clear; they cannot reliably assess a claim whose source, author and execution context are mixed together.

Keep static observations separate from runtime findings

Zeltser's four-stage explanation distinguishes static properties such as strings and hashes from interactive behavioral analysis and manual code reversing. These are different kinds of work. A string resembling a destination does not establish that communication occurred, and a token resembling a capability does not establish implementation or execution. Static properties and behavioral analysis

The distinction applies even when a filename looks suspicious. Describe the artifact you actually have: plaintext, source, a published report or some other permitted material. Do not label a text exercise a Windows executable, infer a platform from an invented filename, or claim that an archive password makes a sample suitable for a recruiter to receive.

For this article's exercise, the input is eleven UTF-8 key/value lines authored locally. Two lines repeat report-demo.invalid/status; three others contain the literal labels autorun, encrypt and collect. Those are teaching strings, not instructions or implemented capabilities. The fixture supplies no process tree, memory capture, packet capture or host telemetry.

The local program reads that text as data. It does not execute the fixture, query a reputation service or contact the marker. Its actual outputs support claims about the text and a deliberate export transformation. They leave every malware-behavior claim untested.

Completed asset one: an identified, limited report packet

The report is titled MP-7 — Completed synthetic text triage and safe-sharing report. Its executive summary states that local parsing found eleven lines, two occurrences of one marker and three capability-like labels. It also states that two demonstration fields were replaced in the exported text and that no malware family, victim or compromise was identified.

The original file has two deliberately fictional fields: an analyst contact at example.invalid and an author-only workspace label. The export replaces their values with [REDACTED_TRAINING_LABEL]. This demonstrates a visible transformation using invented labels; it is not anonymization of a real case or proof that a general redaction process would remove every sensitive item.

Packet component Completed content Boundary recorded
SAFE_REPORT.md Question, executive summary, findings, method, sharing scope and limitations Owned inert text only; no specimen or incident analyzed
manifest.json Original/export names, byte sizes, full SHA-256 values and two transformations Export and original are different artifacts; neither hash is a real malware indicator
analysis.json Eleven lines; marker lines 6–7; three labels on lines 8–10 Counts and positions describe the authored fixture
public-fixture.txt Eleven retained lines with two demonstration values replaced Plaintext teaching export, not a sample download
build_packet.py Complete local parsing, export and hashing implementation Existing Python standard library; no network or specimen execution

The report's completed finding reads: “The supplied text contains one distinct marker twice and three literal capability-like labels. The export replaces two authored demonstration values. These results do not establish network communication, persistence, encryption, collection, malware identity or activity on a host.”

Its provenance statement identifies the author-created source and its relationship to the export. It does not invent a customer, acquisition history or confidential malware collection. Nothing has been published or submitted through a recruiting system; “public” in the export filename describes its intended review scope, not an actual publication event.

The public teaching export's actual SHA-256 is:

6ea0c54ddbf2ddb0ee89059dc698d82feb2397172f7677adc584ea830cc98a47

That digest identifies these particular bytes. It is not a detection recommendation or a value to block in security tools. The source digest is different because the two values changed; the manifest retains both identities instead of carrying the original hash over to the modified export.

NIST's Secure Hash Standard publication describes digests as a way to detect changes relative to previously generated digests. A matching hash does not make the original content truthful or establish that a file is benign. We reviewed the publication page and abstract, not a malware-classification standard. NIST FIPS 180-4 publication

Show the actual repeat check and its limits

We ran the owned builder twice in fresh local Python processes. Both actual runs used CPython 3.9.6 and produced matching analysis, manifest and claim-matrix hashes; their receipts are retained separately. The source bytes remained unchanged. The package records the input/output hashes and actual run receipts; no external software, sandbox or reverse-engineering tool was installed.

From the saved demonstration directory, the commands are:

python3 build_packet.py
python3 review_packet.py

The builder uses hashlib.sha256 over file bytes. Python's current documentation describes that interface; its current documentation version is separate from the older local interpreter we actually tested. We do not claim that every Python version or every input format was verified. Python hashlib documentation

The reproducibility result is narrow: the same owned input and implementation produced the same saved outputs in these two runs. It is not a benchmark, evidence of malware execution or a guarantee that a parser can safely process arbitrary untrusted formats. The script is designed for this declared teaching schema.

Keep the original and derived files named separately. Name the source, export and report revision so a reviewer can trace an image or finding to the right bytes. If a report changes, update its own identity and change record rather than describing every file with the sample's hash.

An analyst corrects an unsupported line in a research notebook while keeping untested questions visible. Editorial illustration.

Completed asset two: a claim review that rejects overstatement

The separate evidence-boundary matrix is a completed review of six specific statements, not an empty checklist. It asks whether each statement is supported by the actual text or transformation, remains untested, or has not been assigned. The reviewer implementation does not import the builder; it independently reads the saved artifacts and checks literal expected lines and hashes.

Proposed statement Completed classification Evidence and missing basis
The marker appears twice in the text Supported static observation Literal lines 6 and 7; one unique marker
The artifact communicated with that marker Untested No network capture, execution or connection observation
The artifact persists at startup Untested autorun is a label; no code or execution establishes persistence
The artifact encrypts or collects data Untested Two labels do not establish either capability
Two demonstration fields were replaced in the export Supported transformation Lines 4 and 5 match the declared replacements
A real malware family or compromise was identified Unassigned No malware sample, family evidence or real incident exists in the exercise

The independent checker passed the saved packet. It also rejected two deliberately altered in-memory controls: a manifest with the wrong export hash, and a matrix that promoted the communication claim to supported. The altered hash and promoted claim existed only in memory; the saved source, export and matrix remained unchanged. This tests whether the finite review catches those two specific errors; it is not an automated determination of arbitrary malware-report quality.

The finished review note says: “Retain two supported statements: literal marker presence and the demonstrated export transformation. Keep three behavior statements untested and the family/compromise statement unassigned. Do not replace missing evidence with a confidence percentage.”

This changes the portfolio wording. “Demonstrated startup persistence and encrypted collection” would overstate the work. A supported alternative is: “Built a reproducible report packet over owned inert text; preserved source/export identity, demonstrated two synthetic redactions and independently rejected a string-to-network-behavior overclaim.”

The latter describes a completed reporting and review task. It does not claim the deeper skills a malware-analysis role may require. If the posting asks for disassembly, unpacking or runtime analysis, record those as separate evidence requirements instead of treating this small exercise as a substitute.

Give a recruiter the report, not an unexplained payload

Lead the review packet with its scope, readable report and artifact index. State which files are included and why. Provide enough readable evidence for review without asking the recipient to disable protections or obtain a specimen. This guide's deliverables are text, Markdown, JSON and original analysis/review scripts, not a malicious executable.

If you later work with authorized real material, verify the applicable sharing permissions and remove details you are not permitted to disclose. A public source does not automatically authorize redistributing every linked artifact. Keep published observations attributed to their publisher, and explain exactly what your own work adds.

A report can be inconclusive and still demonstrate disciplined communication. Preserve the missing evidence beside the finding so the reader sees where interpretation stops. Do not call untested behavior “absent,” and do not turn absence of a demonstrated capability in this text fixture into a claim that an unknown program is safe.

The separate security-analyst portfolio guide concerns log events and investigation hypotheses. This exercise concerns specimen-report identity, safe review packaging and limits on capability claims. Compare a role in LandOffer's recent listings with your report's actual methods. Review the requested format and sharing permissions before including the packet in an application.

Download the worked example files

Download the curated educational companion. Read PUBLICATION-README.md after extraction for the public file layout, replay commands and requirements. The listed offline Python checks were rerun from the extracted archive in a disposable directory. The earlier local report and commands describe the original evidence environment; use the companion README for the portable layout. No whole evidence directory, credentials or compiler binaries are included.

Sources and Further Reading

Public sources reviewed October 23, 2026. All fixture content and resulting hashes are author-created teaching material, not real indicators. No malware, real target, private case or network endpoint was accessed.