Перейти к содержимому
New

DeepSeek V4.1 Flash free for 7 days: 10M tokens a day after a top-up of $10 or more. Until Nov 30.

How it works
aeondave/forensic-techniqueDFIR и threat hunting

forensic-technique

⁨Incident forensics: preservation, triage, timelines, artifact correlation, and reporting across disk images, memory, PCAP, and endpoint/SaaS audit logs.⁩

v2026.11.13-21cd5683 установки+3 за 7 дней39 ЗвездыMITсинхронизировано 10 часов назад

Установка

Ещё нет NotHoly CLI? Сначала установите её.

$ curl -fsSL https://notholy.com/install.sh | sh
$ notholy skills install aeondave/forensic-technique

Пишет в ⁦~/.notholy/skills⁩ после проверки подписи и SHA-256.

ПодписьПодписано, ed25519
aZ6MyASqsz…51WLcYAw==CLI проверяет SHA-256 и подпись, прежде чем записать хоть один файл.
Пакет6d7fedc8…c29b4b7b
6d7fedc863cd…5eacc29b4b7b11 файлов · ⁦24.1 KB⁩
ПроверкаСредний2 замечания
  • injection.no_confirmationreferences/timeline-correlation.md:13
  • Исполняемый скрипт в составе навыкаscripts/forensic_triage.py
Вся проверка
Источникgithub.com/AeonDave/malskill/tree/main/offensive-techniques/forensic-technique@21cd568

Эта запись еще не переведена на русский, поэтому показан текст на другом языке: English.

Goal: produce defensible, reproducible findings from heterogeneous evidence.

When this technique applies

  • Potential compromise requiring reconstruction of user/system activity.
  • Evidence includes disk images, ISO media, RAM dumps, PCAP, endpoint logs, or SaaS audit records.
  • Need to correlate endpoint, filesystem, and network timelines.
  • Need report-ready output for incident response, legal, or internal review.

Boundary with offensive-tools

This skill explains methodology and decision flow. Tool-specific syntax belongs to offensive-tools/* skills.

Initial triage

Before extracting deeply, classify the evidence mix and decide what source can answer the objective fastest with the least interpretive risk.

  • Starting state: is the case driven by disk, memory, PCAP, ISO/media, logs, or a mixed evidence set?
  • First questions: what is the investigation objective, which artifacts are most volatile or most probative, and what timeline window and provenance constraints apply?
  • Immediate actions: verify integrity, normalize time and scope, then inspect the highest-value source first instead of carving everything.
  • Tool-family direction: use forensic acquisition/timeline families (sleuth-kit, autopsy, ftk-imager) for disk/media, volatility3 for memory, and zeek/tcpdump/wireshark for network evidence; use yara and capa only after suspicious artifacts are isolated.
  • Escalation rule: require corroboration across sources before claiming compromise, execution, persistence, or exfiltration.

Agent operating model

The agent should keep this loop:

  1. Preserve and scope evidence.
  2. Prioritize by volatility and investigative value.
  3. Examine source-specific artifacts.
  4. Correlate across sources into one timeline.
  5. Validate claims with independent corroboration.
  6. Report facts, interpretation boundaries, and recommendations.

Do not move to deep extraction/carving before timeline framing and hypothesis definition.

Core forensic model

  1. Preserve: isolate evidence, hash, and record provenance needed for reproducibility.
  2. Collect: acquire additional required artifacts with minimal alteration.
  3. Examine: parse artifacts by source type (disk, memory, network).
  4. Correlate: build a unified timeline across sources.
  5. Conclude: state what is supported by evidence vs assumptions.
  6. Report: produce concise, reproducible findings and follow-up actions.

Evidence-first prioritization

Use volatility and value to decide acquisition/examination order.

  1. Live volatile artifacts (RAM, active connections, running processes)
  2. Network evidence at risk of rotation (live captures, ephemeral logs)
  3. Endpoint high-value non-volatile artifacts (event logs, browser data, registry)
  4. Full disk image deep examination
  5. Enrichment sources (threat intel, external context)

Objective-driven workflows

1. Disk image investigation (E01/DD/RAW)

Use when you need file-system artifacts, deleted-content recovery, execution traces, and persistence evidence.

  • Start with integrity + partition mapping.
  • Enumerate file systems and prioritize user/profile, startup, logs, scheduled tasks, shell history.
  • On Windows evidence, prioritize execution artifacts such as Prefetch, LNK, Jump Lists, AmCache, ShimCache, SRUM, PowerShell logs, services, and scheduled tasks.
  • Build an initial timeline before deep carving.
  • Recover/decode only artifacts linked to hypotheses (do not carve everything by default).
  • Validate notable findings in at least one secondary source (e.g., link file + event log + registry).

Primary tool families:

  • offensive-tools/forensic/sleuth-kit/
  • offensive-tools/forensic/autopsy/
  • offensive-tools/forensic/ftk-imager/
  • offensive-tools/forensic/yara/
  • offensive-tools/forensic/capa/ — capability detection on extracted executables (identifies malware families, TTPs, embedded shellcode)

2. ISO media investigation

Use when the evidence is installer-like media, archives, or mounted image content.

  • Verify hash and mount read-only.
  • Inventory executable/script content and autorun-related metadata.
  • Extract Office/PDF/archive payload chains, embedded objects, external template references, shortcut chains, and script launchers before any dynamic handling.
  • Compare bundle contents against expected vendor structure and signatures.
  • Extract suspicious binaries/configs for static triage and rule scanning.
  • Correlate ISO-delivered artifacts with host execution traces from endpoint evidence.

Primary tool families:

  • offensive-tools/forensic/sleuth-kit/
  • offensive-tools/rev/binwalk/
  • offensive-tools/forensic/yara/

3. PCAP/network forensics

Use when reconstructing communication, exfiltration paths, C2 patterns, or lateral movement.

  • Build traffic summary first (top talkers, protocols, unusual destinations).
  • Pivot by DNS, TLS metadata, HTTP objects, and transfer channels.
  • Identify sequencing: initial access indicators, staging, beaconing, bulk transfer.
  • Tie network sessions to endpoint process/user context when possible.

Primary tool families:

  • offensive-tools/forensic/zeek/
  • offensive-tools/forensic/tcpdump/
  • offensive-tools/network/wireshark/

4. Memory-centered investigation

Use when malware is fileless/injected, or when disk evidence is incomplete.

  • Enumerate processes/parents, command lines, sockets, loaded modules.
  • Hunt for injected regions and anomalous memory protections.
  • Extract process artifacts and correlate with on-disk binaries and PCAP events.
  • Treat memory findings as high-confidence for “what was running now,” then confirm persistence on disk.

Primary tool families:

  • offensive-tools/forensic/volatility3/
  • offensive-tools/forensic/yara/
  • offensive-tools/forensic/capa/ — classify extracted process dumps or unpacked binaries against malware capability ruleset

5. Mixed-source incident reconstruction

Use when you have disk + PCAP + memory and need a single chronology.

  • Create source-local timelines first.
  • Normalize to a single timezone and clock offset model.
  • Merge into one chronology with confidence tags:
    • Direct evidence (observed in source)
    • Corroborated evidence (confirmed by second source)
    • Inference (plausible but unconfirmed)
  • Explicitly record competing explanations for ambiguous events.

6. Log-centric and objective-driven investigations

Use when evidence is mostly EVTX, registry hives, MFT snapshots, API traces, or SaaS audit records, including records inspected through a SIEM UI.

  • Build an objective-to-artifact map first (which artifact can answer each objective decisively).
  • Prefer deterministic extraction over broad hunting (exact key, exact event id, exact stream, exact record).
  • Interpret coded values using the provider's documentation for the full field path and applicable schema/version. Do not transfer enum meanings between fields.
  • After each UI row selection or query change, verify the raw record matches the intended event id and the relevant host/user before saving or interpreting it. A cleared loading indicator does not prove the displayed row is the intended record; discard mismatched extracts and reacquire.
  • For Windows log-heavy cases, prioritize: PowerShell ScriptBlock (4104), process creation (4688/Sysmon 1), service/task creation, Defender/AV alerts, and relevant registry keys. When an ETW trace (.etl) is provided, treat provider name + event id as the primary pointer (e.g. Microsoft-Windows-PowerShell, Microsoft-Antimalware-Scan-Interface, Microsoft-Windows-Kernel-Process).
  • For Linux hosts, prioritize journalctl output from /var/log/journal/* (auth, sshd, sudo, systemd unit start/stop) and /var/log/audit/audit.log (auditd) for structured process/command evidence.
  • For macOS hosts, use the Unified Log system (log collect, log show --predicate ..., or an .logarchive bundle) for process, network, and security signals; classic /var/log text logs are largely deprecated.
  • For "who was logged on / usage timeline" questions, reconstruct sessions from Security 4624 logon types and pair logon/logoff by Logon ID; see references/windows-execution-indicators.md.
  • For API trace cases, reconstruct sequence by API dependency chain (enumeration → allocation/write → execution).
  • Mark each conclusion as direct artifact fact or inferred interpretation. Scenario descriptions and threat intelligence guide hypotheses; they do not establish activity absent from the inspected records.

Primary tool families:

  • offensive-tools/forensic/chainsaw/ — fast triage across EVTX, $MFT, ShimCache, SRUM, and registry hives with Sigma and built-in detection rules; produces multi-source execution timelines

For cloud/SaaS audit records (AWS CloudTrail, Microsoft 365/Entra unified audit log, Google Workspace, Okta), load references/cloud-saas-audit-logs.md for per-provider record/field semantics and event-time vs ingestion-time handling.

Practical quality gates

  • Provenance and hash records exist for all primary evidence.
  • At least one key claim is corroborated across two independent sources.
  • Timeline includes both malicious and benign context to avoid narrative bias.
  • Every IOC or behavioral claim has exact source pointer (artifact + timestamp).
  • Report distinguishes fact, interpretation, and recommendation.

Anti-patterns

  • Starting with deep carving before timeline and hypothesis framing.
  • Treating one artifact as conclusive without corroboration.
  • Mixing local-time and UTC evidence without explicit normalization.
  • Writing conclusion-first reports and backfilling evidence.

Required deliverables from the agent

  1. Evidence inventory with integrity state.
  2. Source-specific findings with exact pointers.
  3. Correlated timeline with confidence labels.
  4. High-confidence claims vs inference boundary.
  5. Report-ready conclusions and next actions.

Resources

  • references/evidence-preservation.md — Load for provenance, hashing, read-only workflow, and evidence-handling order.
  • references/disk-and-iso-analysis.md — Load for filesystem/media triage, startup paths, selective extraction, and validation.
  • references/windows-execution-indicators.md — Load for Prefetch, LNK, Jump Lists, AmCache, ShimCache, SRUM, execution artifacts, user logon/session reconstruction, and timestamp-integrity (timestomping) checks.
  • references/document-malware-analysis.md — Load for Office, PDF, archive, embedded-object, and external-template payload chains.
  • references/pcap-analysis.md — Load for protocol triage, session reconstruction, transfer paths, and network chronology.
  • references/memory-analysis.md — Load for process/module/socket triage, injected regions, transient secrets, and disk/network correlation.
  • references/timeline-correlation.md — Load for timestamp-field semantics, timestamp-integrity checks, clock normalization, data-reduction/pivot workflow, and merging source-local timelines with confidence labels.
  • references/cloud-saas-audit-logs.md — Load for AWS CloudTrail, Microsoft 365/Entra, Google Workspace, and Okta audit-record and timestamp-field semantics.
  • references/lessons-learned-patterns.md — Load for mixed-artifact heuristics and fast-answer investigation patterns.
  • scripts/forensic_triage.py — Run to classify an evidence file and get the first forensic workflow/tool-family recommendation.

Файлы этой версии

Права

ПравоЗаявленоНайдено проверкой
Читает файлыНетНет
Пишет файлыДаДа
Выполняет команды shellДаДа
Ходит в сетьНетНет
Использует коннекторыНетНет
Управляет браузеромНетНет

Замечания проверки

Проверено 13 нояб. 2026 г.

  1. СредняяPrompt injection · injection.no_confirmationreferences/timeline-correlation.md:13
  2. СредняяФайл · Исполняемый скрипт в составе навыкаscripts/forensic_triage.py

Чистая проверка снижает риск, но не доказывает, что пакет безопасен. Читайте то, что ставите.

ВерсияОпубликованаКоммитSHA-256ПодписьИзменение прав
v2026.11.13-21cd56813 нояб. 2026 г.21cd5686d7fedc8Подписана–

Спрашивайте обо всём, что позволяет ваш скоуп.

Создайте аккаунт, пополните баланс и задайте первый вопрос в чате или в терминале.

  1. 1/3

    Создайте аккаунт

    Личный или для компании. Позже можно перейти на аккаунт компании без потери баланса и ключей.

    ЛичныйКомпания
  2. 2/3

    Пополните баланс

    Картой. Оплата за токены, без подписки.

    $5.00минимальное пополнение
  3. 3/3

    Спрашивайте

    В чате, или установите CLI и выполните notholy login в терминале.

    $ notholy login