Skip to content
Free lab

Forensics Lab

An investigation is not a theory. It is what you can prove from the artifacts, and nothing more. Work 10 cases, from insider theft over USB to a phishing ISO that ends in ransomware, a hijacked cloud admin and a browser extension that turned. Log the evidence into custody, tag every artifact, say exactly what it proves, build the timeline, and answer what, when, how and scope. Your lead investigator reviews every finding as you log it.

  • Prefetch
  • Amcache
  • ShimCache
  • Run keys
  • LNK
  • Jump lists
  • Event logs
  • Shellbags
  • SRUM
  • Browser history
  • $MFT
  • USB history
  • Memory
  • Cloud audit
No download, no card
0s

Open the case docket

You are the examiner. Each case gives you a seeded set of artifacts from disk, memory and the cloud. Log the evidence into custody, tag every artifact as evidence or noise, say exactly what it proves, build the timeline, then answer what, when, how and scope. Your lead investigator, Dana Okafor, reviews every finding as you log it.

10
Investigations
17
Artifact types
3
Rounds per docket
75%
To clear

Pro docket: rounds climb in case level and noise, starting at more noise. Scoring: +1 per evidence item logged, +1 if your claim about it is right, +1 for its timeline phase, +2 per question, +1 per custody check, +0.5 per red herring ruled out; −1 per red herring cited and per hint.

A new seed is drawn when you start

Or work one case on its own

Artifact field guide: what each one proves, and what it does not+

Prefetch

C:\Windows\Prefetch\<EXE>-<hash>.pf

Proves: The executable ran on this host: run count, up to 8 last-run times (Win8+), and files it touched in its first seconds.

Does not prove: Which user ran it. Prefetch is disabled by default on Windows Server, so no .pf file there is not proof that nothing ran.

Amcache

C:\Windows\AppCompat\Programs\Amcache.hve

Proves: A file was present on the host, with its path and SHA-1 (of the first ~31 MB), and when it was inventoried.

Does not prove: Execution, on its own: on Windows 10/11 an InventoryApplicationFile entry records presence. Corroborate with Prefetch, EDR or 4688.

ShimCache (AppCompatCache)

SYSTEM hive \ControlSet001\Control\Session Manager\AppCompatCache

Proves: The OS saw this file path. The timestamp is the file's last-modified time ($SI), and entries are written to the hive at shutdown.

Does not prove: Execution on Windows 10/11 (there is no execution flag), and the timestamp is not a run time.

Registry Run keys

NTUSER.DAT / SOFTWARE \Microsoft\Windows\CurrentVersion\Run (and RunOnce)

Proves: Something is configured to start at logon (persistence), and the key's LastWrite time bounds when it was set.

Does not prove: That the program ran, or ran successfully. A Run value is configuration, not execution.

LNK shortcut files

%APPDATA%\Microsoft\Windows\Recent\*.lnk

Proves: This user profile opened the target file or folder. The LNK keeps the target path, the target's times, drive type and volume serial number.

Does not prove: That the file was copied, edited or sent anywhere. It also does not keep the file's contents.

Jump lists

%APPDATA%\Microsoft\Windows\Recent\AutomaticDestinations\<AppID>.automaticDestinations-ms

Proves: This user opened these files with that application (identified by AppID), with access times.

Does not prove: That the application was used to send or exfiltrate the files.

Windows event logs

C:\Windows\System32\winevt\Logs\*.evtx

Proves: Specific OS events: logons and their type, account and group changes, service and task creation, process creation (4688, command line only if enabled), log clearing.

Does not prove: Anything the audit policy was not configured to record. Silence in a log is not evidence of absence.

Shellbags

UsrClass.dat \Local Settings\Software\Microsoft\Windows\Shell\BagMRU

Proves: This user browsed this folder in Explorer, including folders on removable drives or since deleted.

Does not prove: That any file inside the folder was opened or copied.

SRUM

C:\Windows\System32\sru\SRUDB.dat

Proves: Bytes sent and received per application and user, aggregated over roughly hourly intervals.

Does not prove: Where the data went or what it was. SRUM has no destination, URL or file names.

Browser history

Chrome/Edge: %LOCALAPPDATA%\...\User Data\Default\History (SQLite: urls, visits, downloads)

Proves: This browser profile visited the URL or downloaded the file at that time, with referrer and download target path.

Does not prove: Who sat at the keyboard, or that a downloaded file was opened. Private windows are not recorded.

$MFT timestamps

NTFS $MFT record: $STANDARD_INFORMATION ($SI) and $FILE_NAME ($FN) attributes

Proves: MACB times for a file. $SI can be changed from user mode (timestomping); $FN is normally set by the kernel, so $SI earlier than $FN, or $SI with zeroed sub-seconds, is a tell.

Does not prove: Who created the file or whether it ran.

USB device history

SYSTEM \Enum\USBSTOR, MountedDevices, setupapi.dev.log, Partition/Diagnostic 1006

Proves: A specific device (vendor, product, serial) connected, with first-install, last-arrival and last-removal times and the drive letter it got.

Does not prove: That any file was copied to or from it. Tie it to LNK/shellbag volume serials and paths for that.

Memory image

RAM capture (process list, command lines, network connections, injected regions)

Proves: The state at capture time: running processes and parents, command lines, open connections, private executable memory.

Does not prove: History before the capture or persistence across reboots. It is volatile and only what was resident.

EDR telemetry

Endpoint agent console: process tree, command lines, network and file events

Proves: Process lineage, command lines and connections as the agent recorded them, with times.

Does not prove: What happened while the agent was absent or blind, or the content of the traffic.

Cloud audit log

Control-plane audit trail (identity sign-ins, API calls) exported from the tenant

Proves: An identity or session made these calls from this IP and user agent, with parameters and result.

Does not prove: Which human was behind the session. Data-plane reads (object downloads, mailbox reads) appear only if those logs were enabled.

Mail trace

Mail gateway / message trace and mailbox audit

Proves: Message metadata: sender, recipients, subject, time, URLs and attachments seen, and mailbox rule changes.

Does not prove: That a recipient read or clicked, unless click or read events were logged.

File on disk

Files and configuration collected from the image

Proves: What the file contains and its timestamps at acquisition.

Does not prove: Who wrote it or when it was used, without other artifacts.

TL;DR

What this lab is

5 points
  • A free digital-forensics (DFIR) lab: 10 realistic investigations, from a departing engineer copying designs to USB to a phishing ISO that ends in ransomware, a hijacked cloud admin and a browser extension that turned malicious.
  • Each case hands you a seeded set of artifacts across 17 types: Prefetch, Amcache, ShimCache, Run keys, LNK files, jump lists, event logs, shellbags, SRUM, browser history, $MFT timestamps, USB history, memory and cloud audit logs.
  • You work it like a real examiner: log the evidence into chain of custody, tag every artifact as evidence or noise, say exactly what it proves, build the timeline, then answer what, when, how and scope.
  • Every finding gets feedback from your lead investigator as you log it, including what that artifact does not prove (Prefetch has no user, ShimCache has no run time, USB history has no file list).
  • Dockets of 3 rounds climb in difficulty and are seeded, so a ?seed= link replays the same case. The debrief shows your timeline and report on the page; Pro: download as Markdown.
Examining evidence and keeping the chain of custody. AI-generated; the people and organisations shown are fictional.

Digital forensics training: what to learn and in what order

  1. Evidence handling: chain of custody and why you never work on the original.
  2. Host artifacts: what Prefetch, ShimCache, LNK files and event logs each prove, and what they do not.
  3. Timelines: combine artifacts into one sequence and mark the gaps.
  4. Network and cloud evidence: packet captures, DNS and cloud audit logs.
  5. Reporting: state findings, confidence and limits in plain language.

Related reading: malware analysis for beginners, incident response training and how to learn cybersecurity.

Frequently asked questions

What is the Forensics Lab?

A free, in-browser digital forensics and incident response (DFIR) training lab. You work 10 fictional investigations from the artifacts: you log the evidence into chain of custody, tag each artifact as evidence or a red herring, state what it proves, build a timeline by phase, and answer the four questions every investigation report must answer: what happened, when, how, and the scope.

Which Windows artifacts does it cover?

Prefetch, Amcache, ShimCache (AppCompatCache), registry Run keys, LNK shortcut files, jump lists, Windows event logs (4624, 4625, 4688, 4698, 4720, 4732, 7045, 1102, Sysmon and Partition/Diagnostic 1006), shellbags, SRUM and $MFT timestamps, plus USB device history, browser history, memory images, EDR telemetry, mail traces and cloud audit logs.

Why does it ask what each artifact proves?

Because the most common mistake in forensics is claiming more than an artifact shows. Prefetch proves a program ran but not which user ran it. ShimCache records a path, and its timestamp is the file modified time, not a run time. USB history proves a device connected, not that files were copied. SRUM gives bytes per app, not destinations. The lab scores the claim, not just the pick.

How does it teach timestomping?

Several cases include files whose $STANDARD_INFORMATION times were forged. You learn the tells: $SI earlier than $FILE_NAME, zeroed sub-second values, and MFT record numbers sitting among recently created files. A decoy shows the classic false positive, where an installer legitimately sets an old $SI time.

What is the seed, and how do rounds work?

A docket is 3 rounds that climb in case level and noise. Rookie starts with light noise, Pro and Elite start harder. The seed decides which cases appear, which red herrings are mixed in and the order of options, so sharing a link with ?seed= gives a colleague exactly the same round. You can also pick any single case.

Is the report export free?

No. Download is Pro. The lab and the on-page debrief are free to play and read; Pro exports your chain-of-custody notes, timeline, findings, ruled-out items and each artifact's limitations as a Markdown file.

Are the cases real?

No. Every person, company, host, domain (.example) and IP address (documentation and private ranges) is invented. The artifact behaviour is real Windows and cloud behaviour, written to be accurate. The AI coding-agent case follows the endpoint artefacts described by Invictus Incident Response, Frontier Forensics: Anthropic (Oct 2026).

What is network forensics?

Network forensics is investigating an incident from network evidence: packet captures, flow records, DNS and proxy logs and firewall logs. It answers who talked to whom, when, and how much data moved. This lab focuses on host and cloud artifacts; pair it with packet-capture practice, as described in the malware analysis guide, for the network side.

What does digital forensics training involve?

Digital forensics training teaches evidence handling and chain of custody, how to read artifacts from operating systems, browsers, memory and cloud logs, how to build a timeline, and how to report only what the evidence proves. A good course or course-like practice has graded hands-on cases and a written report at the end, which this lab provides for free in the browser.

Where can I practise digital forensics for free?

This lab runs fictional cases in your browser with nothing to install. For a wider path, see the hands-on training guide and the free course library, and for the response side use the IR Playbook Lab.

The tools people use most on PlayCISO.

All tools