Windows Host Case: answer key
Every question with its answer and the reasoning. Work the drill first, or the key only tells you what, not how.
Stages: Event logs, Registry, Timeline, Collection.
Event logs
-
Triage this raw Windows log for Event ID 4624 ("An account was successfully logged on", channel: Security). Read it and type what was actually recorded for each field below.
An account was successfully logged on. Subject: Security ID: S-1-5-18 Account Name: WORKSTATION01$ Account Domain: CORP Logon ID: 0x3E7 Logon Information: Logon Type: 3 Restricted Admin Mode: - Virtual Account: No Elevated Token: No Impersonation Level: Impersonation New Logon: Security ID: S-1-5-21-3623811015-3361044348-30300820-1104 Account Name: jsmith Account Domain: CORP Logon ID: 0x3E7F2A1 Linked Logon ID: 0x0 Network Account Name: - Network Account Domain: - Logon GUID: {00000000-0000-0000-0000-000000000000} Process Information: Process ID: 0x0 Process Name: - Network Information: Workstation Name: WORKSTATION01 Source Network Address: 203.0.113.10 Source Port: 49234 Detailed Authentication Information: Logon Process: NtLmSsp Authentication Package: NTLM Transited Services: - Package Name (NTLM only): NTLM V2 Key Length: 128Answer: Logon Type: 3 · Account Name (Subject): WORKSTATION01$ · Account Name (New Logon): jsmith · Source Network Address: 203.0.113.10
Event ID 4624: "An account was successfully logged on." Records that a logon session was created on the computer being accessed. Captures the account (Subject and New Logon), the Logon Type (Interactive, Network, RemoteInteractive/RDP, Service, Batch, Unlock, NewCredentials, etc.), the process that requested the logon, and the source workstation/IP. The foundational record of successful authentication on a host. It is the starting point for reconstructing who logged on where and how, correlating lateral movement across a network, and building an authentication timeline during an investigation. The Logon ID field ties it to related events like 4634, 4647, 4672, 4627, and to 4769 on a domain controller via the Logon GUID.
Read more: Full reference: Event ID 4624
-
Triage this raw Sysmon log for Event ID 11 ("FileCreate", channel: Microsoft-Windows-Sysmon/Operational). Read it and type what was actually recorded for each field below.
File created: RuleName: - UtcTime: 2024-01-15 14:42:33.012 ProcessGuid: {A1B2C3D4-E5F6-4A5B-8C9D-0E1F2A3B0006} ProcessId: 7788 Image: C:\Users\jsmith\AppData\Local\Temp\invoice.exe TargetFilename: C:\Users\jsmith\AppData\Roaming\update.dll CreationUtcTime: 2024-01-15 14:42:33.012 User: CORP\jsmithAnswer: Image: C:\Users\jsmith\AppData\Local\Temp\invoice.exe · TargetFilename: C:\Users\jsmith\AppData\Roaming\update.dll · CreationUtcTime: 2024-01-15 14:42:33.012
Event ID 11: "FileCreate." Logs file creation and overwrite operations, including the target path and the process that performed it. One of the highest-value events for catching malware dropping payloads into autostart locations, temp/download directories, or other common staging paths during initial infection.
Read more: Full reference: Event ID 11
-
Which MITRE ATT&CK technique ID does Event ID 4720 ("A user account was created", channel: Security) most commonly map to? (e.g. "T1078")
Answer: T1136
Marks the starting point for tracking a new identity in the environment. Unexpected or off-hours account creation, especially by a non-IT account, is a classic indicator of an adversary establishing persistence. This maps to Create Account (T1136). See the full technique writeup on the ATT&CK Coverage Map for detection guidance and mitigations.
Read more: Full reference: Event ID 4720
-
Triage this raw Windows log for Event ID 1102 ("The audit log was cleared", channel: Security). Read it and type what was actually recorded for each field below.
The audit log was cleared. Subject: Security ID: S-1-5-21-3623811015-3361044348-30300820-500 Account Name: administrator Domain Name: CORP Logon ID: 0x3E7F2A1
Answer: Account Name (Subject): administrator · Domain Name: CORP · Logon ID: 0x3E7F2A1
Event ID 1102: "The audit log was cleared." Generated by the Windows Eventlog service itself (Provider Microsoft-Windows-Eventlog, not the Security-Auditing provider) every time the Security event log is cleared. The event's LogFileCleared payload records the SID, account name, domain, and logon ID of whoever performed the clear. Because it's written by the Eventlog service as part of the clear operation, 1102 is the canonical 'someone cleared the Security log' tripwire and normally shouldn't occur at all in routine operation. Microsoft's own guidance is to monitor and investigate every occurrence. Forwarding it off-host / alerting on it in real time matters because it can otherwise be lost along with the rest of the log it documents.
Read more: Full reference: Event ID 1102
-
Triage this raw Sysmon log for Event ID 22 ("DNSEvent (DNS query)", channel: Microsoft-Windows-Sysmon/Operational). Read it and type what was actually recorded for each field below.
Dns query: RuleName: - UtcTime: 2024-02-11 09:43:12.204 ProcessGuid: {a1b2c3d4-1234-65a1-0000-001122334466} ProcessId: 7148 QueryName: update.example.com QueryStatus: 0 QueryResults: ::ffff:198.51.100.42; Image: C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exeAnswer: QueryName: update.example.com · Image: C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe · QueryStatus: 0 · QueryResults: ::ffff:198.51.100.42;
Event ID 22: "DNSEvent (DNS query)." Logs DNS queries made by a process, whether the result succeeded, failed, was cached, or not. Telemetry for this event was added for Windows 8.1 and is not available on Windows 7 and earlier. Per-process DNS query logging helps identify malware beaconing to C2 domains, DNS tunneling, or domain-generation-algorithm activity by tying specific lookups back to the requesting process.
Read more: Full reference: Event ID 22
-
Which MITRE ATT&CK technique ID does Event ID 10 ("ProcessAccess", channel: Microsoft-Windows-Sysmon/Operational) most commonly map to? (e.g. "T1078")
Answer: T1003.001
This is the primary Sysmon telemetry for detecting credential-dumping tools reading LSASS memory and other process-injection precursors, by examining the target image and requested access rights (GrantedAccess). This maps to LSASS Memory (T1003.001). See the full technique writeup on the ATT&CK Coverage Map for detection guidance and mitigations.
Read more: Full reference: Event ID 10
-
Triage this raw Windows log for Event ID 4625 ("An account failed to log on", channel: Security). Read it and type what was actually recorded for each field below.
An account failed to log on. Subject: Security ID: S-1-0-0 Account Name: - Account Domain: - Logon ID: 0x0 Logon Type: 3 Account For Which Logon Failed: Security ID: S-1-0-0 Account Name: jsmith Account Domain: CORP Failure Information: Failure Reason: Unknown user name or bad password. Status: 0xC000006D Sub Status: 0xC000006A Process Information: Caller Process ID: 0x0 Caller Process Name: - Network Information: Workstation Name: WORKSTATION01 Source Network Address: 203.0.113.15 Source Port: 49523 Detailed Authentication Information: Logon Process: NtLmSsp Authentication Package: NTLM Transited Services: - Package Name (NTLM only): - Key Length: 0
Answer: Logon Type: 3 · Account Name (Account For Which Logon Failed): jsmith · Sub Status: 0xC000006A · Source Network Address: 203.0.113.15
Event ID 4625: "An account failed to log on." Logged for any failed logon attempt, regardless of the reason, on the computer where the attempt was made. Captures the target account, the Logon Type attempted, the source workstation/IP, and a Status/Sub Status code identifying the specific failure reason. The central event for detecting brute-force attacks, password-spraying, account enumeration, and account lockouts. The Status/Sub Status codes let an analyst distinguish a simple mistyped password from a disabled, locked-out, or expired account, or a logon-hours/authorized-workstation violation.
Read more: Full reference: Event ID 4625
-
Triage this raw Windows log for Event ID 4672 ("Special privileges assigned to new logon", channel: Security). Read it and type what was actually recorded for each field below.
Special privileges assigned to new logon. Subject: Security ID: S-1-5-21-3457937927-2839227994-823803824-500 Account Name: administrator Account Domain: CORP Logon ID: 0x8A2F1 Privileges: SeSecurityPrivilege SeBackupPrivilege SeRestorePrivilege SeTakeOwnershipPrivilege SeDebugPrivilege SeSystemEnvironmentPrivilege SeLoadDriverPrivilege SeImpersonatePrivilege
Answer: Account Name (Subject): administrator · Logon ID: 0x8A2F1 · Privilege granting LSASS memory access (used by tools like Mimikatz): SeDebugPrivilege
Event ID 4672: "Special privileges assigned to new logon." Fires whenever a new logon session is granted one or more sensitive privileges from a fixed list (SeTcbPrivilege, SeBackupPrivilege, SeCreateTokenPrivilege, SeDebugPrivilege, SeEnableDelegationPrivilege, SeAuditPrivilege, SeImpersonatePrivilege, SeLoadDriverPrivilege, SeSecurityPrivilege, SeSystemEnvironmentPrivilege, SeAssignPrimaryTokenPrivilege, SeRestorePrivilege, and SeTakeOwnershipPrivilege), and lists exactly which were assigned. Flags every logon session that carries administrator-equivalent power. Because SYSTEM logons trigger it constantly, analysts typically filter for non-SYSTEM/non-service accounts to spot unexpected privileged sessions: a common signal for privilege escalation or misuse of an admin account.
Read more: Full reference: Event ID 4672
-
Which MITRE ATT&CK technique ID does Event ID 7045 ("A service was installed in the system", channel: System) most commonly map to? (e.g. "T1078")
Answer: T1543.003
Because it's written to the System log by the Service Control Manager itself, it is enabled by default on every Windows system with no advanced audit-policy configuration required, making it one of the most reliable, universally available artifacts for detecting service-based persistence and lateral movement. Classic example: PsExec drops PSEXESVC.exe as a temporary service on the target, generating a 7045 that is a strong indicator of PsExec-based lateral movement. The ImagePath is the primary triage field: services pointing at Temp, AppData, or other non-standard paths are a strong red flag. This maps to Windows Service (T1543.003). See the full technique writeup on the ATT&CK Coverage Map for detection guidance and mitigations.
Read more: Full reference: Event ID 7045
-
Triage this raw Sysmon log for Event ID 3 ("Network connection", channel: Microsoft-Windows-Sysmon/Operational). Read it and type what was actually recorded for each field below.
Network connection detected: RuleName: - UtcTime: 2024-01-15 14:36:02.789 ProcessGuid: {5a8f3c2b-7e41-65b0-1200-000000001b00} ProcessId: 4521 Image: C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe User: CORP\jsmith Protocol: tcp Initiated: true SourceIsIpv6: false SourceIp: 198.51.100.23 SourceHostname: WKSTN-JSMITH.corp.local SourcePort: 51322 SourcePortName: - DestinationIsIpv6: false DestinationIp: 203.0.113.10 DestinationHostname: - DestinationPort: 443 DestinationPortName: httpsAnswer: Image: C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe · DestinationIp: 203.0.113.10 · DestinationPort: 443 · User: CORP\jsmith
Event ID 3: "Network connection." Logs TCP/UDP network connections initiated by a process, including source/destination hostnames, IP addresses, port numbers, and IPv6 status. Disabled by default because of the volume it generates. Ties outbound or inbound network activity to a specific process and ProcessGuid, which is essential for identifying C2 beaconing, lateral movement, or exfiltration channels during an investigation.
Read more: Full reference: Event ID 3
Registry
-
ShimCache on a compromised host lists C:\ProgramData\Intel\IntelSvc.exe. The binary is gone. What does this entry establish?
Answer: The file existed on this system at some point, with the file timestamp shown
ShimCache records that Windows evaluated a file for compatibility shimming. That happens when a file is executed, and it also happens when Explorer merely browses the directory holding it. So the entry proves presence and gives you the file's $STANDARD_INFORMATION last-modified time. It does not prove execution. Writing "ShimCache shows it ran" is the most common Registry error in real reports, and it survives review because it sounds reasonable.
Read more: Registry reference: AppCompatCache
-
Sort each artifact by whether it can support a claim that a program actually ran.
Answer: UserAssist → Proves execution · BAM (Background Activity Moderator) → Proves execution · ShimCache / AppCompatCache → Proves presence only · Amcache → Proves presence only
UserAssist counts GUI launches per user and BAM records last-execution times per user SID, so both support an execution claim. ShimCache and Amcache record that Windows saw the file. Amcache is still valuable because it carries a SHA-1, which identifies a binary that has since been deleted. The defensible pattern is to pair them: presence and hash from one, execution and time from the other.
Read more: Registry reference: execution evidence
-
This UserAssist entry has been decoded. Pull the three facts you would put in a timeline row.
Key : NTUSER.DAT\...\UserAssist\{CEBFF5CD-ACE2-4F4F-9178-9926F41749EA}\Count Value name (ROT13) : P:\CebtenzQngn\Vagry\VagryFip.rkr Decoded name : C:\ProgramData\Intel\IntelSvc.exe Run count : 9 Focus time : 2h 14m 33s Last executed : 2026-03-02T21:47:09ZAnswer: Program path: C:\ProgramData\Intel\IntelSvc.exe · Run count: 9 · Last executed (UTC): 2026-03-02T21:47:09Z
The GUID matters as much as the values. {CEBFF5CD-...} counts executable launches, so this is nine GUI launches of a binary sitting in ProgramData under an Intel-sounding name. The ROT13 on the value name is obfuscation for tidiness rather than security. Note what this cannot tell you: anything launched from a script, a service or a scheduled task never reaches UserAssist at all.
Read more: Registry Value Decoder
-
Match each investigative question to the Registry key that answers it.
Answer: Which physical USB device was attached, by serial number? → USBSTOR · Which folders did this user browse, including deleted ones? → ShellBags · What did the user type into the Run dialog? → RunMRU · Which machines did this user RDP out to? → Terminal Server Client · Which Wi-Fi networks has this machine joined, and when? → NetworkList
Two of these carry a catch worth remembering. A USBSTOR serial ending in "&0" was generated by Windows because the device supplied none, so it is not unique and cannot match the same device on another system. And Terminal Server Client records outbound RDP only; inbound sessions live in the Terminal Services event logs, not the Registry.
Read more: Registry reference: by question
-
You copied SYSTEM, SOFTWARE and every NTUSER.DAT off a running host. Your parser shows nothing from the last two days. What is the most likely explanation?
Answer: The transaction logs were not collected, so recent writes have not been replayed into the hives
A hive on a live system is routinely stale. Recent changes sit in the .LOG1 and .LOG2 transaction logs waiting to be written back, so a copy taken without them silently loses the most recent activity, which is usually the activity you care about. Collect the logs alongside every hive and let the parser replay them. This is also why UsrClass.dat matters: it holds the ShellBags for This PC and removable media that NTUSER.DAT alone will not give you.
-
What does this Winlogon configuration tell you?
Key : SOFTWARE\Microsoft\Windows NT\CurrentVersion\Winlogon Shell : explorer.exe Userinit: C:\Windows\system32\userinit.exe,C:\ProgramData\Intel\IntelSvc.exe
Answer: A second binary has been appended to Userinit and will run at every interactive logon
Userinit should read exactly C:\Windows\system32\userinit.exe with a trailing comma. That comma is legitimate and is not a sign of corruption, which is worth knowing because it looks malformed. Anything appended after it runs at every interactive logon with the user's privileges. Shell is untouched here and correctly reads explorer.exe.
Read more: Registry reference: Winlogon
-
Defender exclusions on a compromised host list C:\ProgramData\Intel\ and the extension .bin. Why is this one of the most useful findings on the box?
Answer: It names the directory and file type the attacker intended to use
An attacker-added exclusion is a map to their own staging area. It tells you exactly where to look and what to look for, before you have found a single sample. Legitimate exclusions do exist for performance reasons, so compare against the organization's baseline rather than treating every entry as hostile. Defender being disabled outright is a different key: DisableAntiSpyware under the policy path.
Read more: Registry reference: defense evasion
-
Put these steps in the order that produces a defensible Registry timeline.
Answer: Collect the hives together with their .LOG1 / .LOG2 transaction logs Read TimeZoneInformation to establish the offset Resolve SIDs to usernames via ProfileList Parse execution artifacts and place them on the timeline
Collection comes first because a hive taken without its logs is already missing the recent end of the story. The timezone comes next: some Registry timestamps are FILETIME in UTC and others, NetworkList among them, are SYSTEMTIME in local time, so a timeline built before you know the offset is confidently wrong by hours. SID resolution turns every later artifact into a name. Only then does parsing produce something you can defend.
-
You need ShellBags showing which removable drives a user browsed. Which hive do you need?
Answer: UsrClass.dat, from the user's AppData\Local\Microsoft\Windows
ShellBags are split across two per-user hives. NTUSER.DAT holds the desktop namespace; UsrClass.dat holds This PC and removable media. Pulling only NTUSER.DAT is a common collection gap and it loses exactly the external-drive evidence that usually matters. Collect both, for every user on the system rather than only the suspect, because attribution changes mid-case more often than people expect.
Read more: Registry reference: where the hives live
-
BAM shows no entry for a binary you believe ran three weeks ago. What is the correct reading?
Answer: BAM keeps roughly a week of history, so its silence says nothing about three weeks ago
BAM is genuinely execution evidence and it is attributed to a user SID, which makes it one of the better artifacts available. It also holds about seven days on most systems and is commonly cleared by a feature update. Absence outside that window is a property of the artifact. The other three options are all possible in principle, but none of them is what the evidence here supports, and saying so is the difference between a finding and a guess.
Read more: Registry reference: BAM
Timeline
-
One Windows host, one intrusion. Put these findings in the order the underlying events happened.
Answer: Sysmon 1: rundll32.exe spawned by winword.exe Zeek dns.log: a lookup for updates.example.net Zeek conn.log: a 6-hour connection with 40 KB out and 900 bytes in System 7045: a service named "IntelSvc" was installed Security 4624 logon type 3 from a workstation in the same subnet
Execution comes first: a document spawning rundll32 is the initial-access moment. The name lookup precedes the connection it resolves for, and the long asymmetric connection is the C2 channel that follows. Persistence is installed once the channel works, not before, because an operator who loses access before establishing it has nothing to persist. Lateral movement is last, because it needs the foothold that everything above it built. The connection direction is the tell on the last one: type 3 INTO this host means it is now a target, not the source.
Read more: Windows Event IDs
-
Match each question to the artifact that actually answers it.
Answer: Which USB device was attached, and when was it first seen? → Windows Registry · What did an account do after escalating with sudo? → Linux · What persistence was installed for one user, running every ten minutes? → macOS · What hostname did the client ask for inside an encrypted session? → Network · Which folders did a user open in Explorer? → Windows Registry · Which service was enabled to restart itself on failure? → Linux
USBSTOR and Explorer's ShellBags are Registry-only. The sudo question is auditd, and the useful field is auid, the login UID that survives escalation. The real UID after sudo says root and identifies nobody. A per-user LaunchAgent plist with StartInterval is the macOS persistence shape. The SNI in a TLS handshake is readable without decrypting anything, which is why ssl.log answers a question about encrypted traffic. And a systemd unit with Restart=always is what makes a Linux service persistent rather than a one-shot.
Read more: ATT&CK Data Sources
-
A single executable, C:\ProgramData\Intel\IntelSvc.exe, runs once. Order these artifacts by the moment each one records.
Answer: $STANDARD_INFORMATION created time on the file Amcache: SHA-1 and first-seen entry Sysmon 1: process creation, with the command line Prefetch: last run time
The file has to exist before anything else can happen to it. Amcache records that Windows cataloged the binary, which happens around installation or first execution and carries a hash you can still use after the file is deleted. Sysmon records the launch itself, with the command line: the richest single record of the four. Prefetch is written by the loader after the process starts, so it lags the process-creation event by a few seconds. Those few seconds matter: a Prefetch time that PRECEDES the Sysmon event is not a fast disk, it is a different execution you have not found yet.
Read more: Registry Forensics Reference
-
Your timeline has a four-hour gap on one host with no events at all. What is the FIRST thing to establish?
Answer: Whether the sources covering that window were collecting, and what their retention is
All four are worth checking, and three of them are conclusions. Coverage is the one that decides which of the others is even a question. A gap in a log that was never enabled, or that has already rolled over, is not evidence of anything, and treating it as anti-forensics puts a claim in a report that the data cannot support. Log clearing leaves its own record (Security 1102, and the audit trail in journald or auditd), so it is checkable rather than inferable.
Read more: ATT&CK Data Sources
-
A phishing case on a macOS laptop. Order what happened, using what each artifact records.
Answer: Quarantine attribute set on ~/Downloads/invoice.pdf Unified log: Gatekeeper assessment for the downloaded file A LaunchAgent plist appears in ~/Library/LaunchAgents/ A TLS session to a host whose certificate was issued two days earlier
The quarantine attribute is written by the downloading application at download time, before the file is opened. Gatekeeper assesses it at first launch, which is the moment the user acted. Persistence is installed after that succeeded. The certificate detail is the trap: it was ISSUED two days earlier, which is infrastructure preparation, but the SESSION is the last event here. Confusing when a certificate was created with when it was used is a common ordering error, and the two can be weeks apart.
Read more: macOS Artifacts
-
Sort each artifact by the strongest claim it can support on its own.
Answer: Prefetch entry with a run count → Proves execution · ShimCache / AppCompatCache entry → Proves presence only · Zeek conn.log record → Proves a connection · auditd record with an auid → Proves an account acted · Amcache entry with a SHA-1 → Proves presence only · Security 4624 with a logon type and account name → Proves an account acted
ShimCache and Amcache are the two that get overclaimed. Both record that Windows saw the file: ShimCache from a compatibility evaluation that browsing a folder also triggers, Amcache from cataloging. Neither establishes execution, and "ShimCache shows it ran" survives review constantly because it sounds reasonable. Amcache is still valuable for its hash, which identifies a binary that has since been deleted. Pair a presence artifact with an execution artifact and the claim becomes defensible.
Read more: Registry Forensics Reference
-
A Linux server. Order these records by when the underlying events happened.
Answer: auth.log: accepted publickey for user deploy from an external address auditd: execve of /usr/bin/wget with an external URL auditd: execve of /bin/sh -c with a base64 -d in the command line auth.log: sudo session opened for user deploy A file appears at /etc/systemd/system/telemetry.service
Access comes first. Download, then decode-and-run the thing that was downloaded. The sudo session sits where it does because writing to /etc/systemd/system needs root, and the unit file appearing is the consequence of that escalation rather than a separate decision. The key field throughout is auid: it stays as deploy's login UID across the sudo boundary, which is what ties the root-owned unit file back to a person. Reading the real UID instead gives you root, which identifies nobody.
Read more: Linux Artifacts
-
Zeek logs for one exfiltration. Order these by when Zeek would have written them.
Answer: dns.log: the query and its answer ssl.log: the handshake, with SNI and JA3 files.log: a reassembled file and its SHA-256 conn.log: the completed connection, with duration and byte counts
conn.log is LAST, and that surprises people. Zeek writes a connection record when the connection ENDS, because duration and byte totals are not known until then. So the log that describes the whole conversation is written after every log describing what happened inside it. This is also why the uid field exists: it is stamped on every protocol record for a connection, and it is what joins the DNS lookup, the handshake, the transferred file and the totals into one story. Filtering each log separately by IP loses that linkage.
Read more: Zeek Log Fields
-
Two events on different hosts are 90 seconds apart in your timeline, and you want to say one caused the other. What has to be true first?
Answer: Both hosts' clocks are reconciled to one reference, and the offset between them is recorded
A 90-second gap means nothing if the two clocks disagree by two minutes, and host clocks routinely do. Reconcile first, record the correction in the report, then reason about sequence. The same account and the same domain are corroboration, not prerequisites. Plenty of causal chains cross accounts, and that crossing is often the finding.
Read more: ATT&CK Data Sources
-
Your timeline is complete and internally consistent. What does that establish?
Answer: That the events you collected are consistent with your account, and nothing about the events you did not collect
A consistent timeline is a strong result and a weak proof. It says the sources you had agree with each other and with your reading of them. It says nothing about the host with no agent, the log that had already rolled, or the action that leaves no artifact anywhere. Reports that state their coverage alongside their conclusions age well. Reports that state only conclusions get revisited when the next host turns up.
Read more: ATT&CK Data Sources
Collection
-
A workstation is suspected of running a backdoor and is still powered on. Order these by what you collect first.
Answer: Established network connections Physical memory Full disk image Firewall and proxy logs for the host
Connections decay in seconds and are frequently the only place a C2 address ever appears, so they go first even though the capture is trivial. Memory is next and is the highest-value source overall. Disk survives until overwritten, so it can wait for hours. The remote logs come last not because they matter least, but because nothing you do to the host changes them.
Read more: Order of volatility
-
A helpdesk technician rebooted the suspect host before you arrived. Which evidence did that specifically destroy?
Answer: Physical memory, and with it injected code, decrypted configuration and cleartext credentials
Memory does not survive a reboot. Everything that only ever existed unpacked in RAM went with it: injected code, the decrypted config a packed sample was using, credentials in process memory, and the connection table. The other three are file-backed and are still there. This is why "have you tried restarting it" is a forensic event, and why it usually happens before anyone calls you.
Read more: Order of volatility: physical memory
-
Where should a memory image be written, and why?
Answer: To removable media or a network share, because writing locally overwrites unallocated space
A memory image is the size of RAM. Writing it to the suspect disk consumes unallocated clusters, which is exactly where deleted files and carvable fragments live, so the act of collecting one source destroys another. Hashing afterwards proves the copy is intact; it does not undo the overwrite. Record where you wrote it, because the next analyst needs to tell your footprint from the attacker's.
Read more: Order of volatility: collection decisions
-
Match each question to the source that actually answers it.
Answer: What was the decrypted C2 configuration? → Physical memory · Which external address is the host talking to right now? → Network connections · When was the persistence key written? → Disk · What did the host do three weeks ago, before it was compromised? → Remote logs · What did the attacker type into a process that has since exited? → Physical memory · Which files were staged in a temp directory and then deleted? → Disk
The last one is worth arguing about. A deleted staging directory is recoverable from unallocated space on disk, provided the clusters have not been reused, which is a race you lose slowly rather than instantly. And "three weeks ago" belongs to remote logs because host-based evidence of that age is usually gone or altered, while the logs the attacker could not reach are not.
-
A production database server is suspected of exfiltrating data. It cannot be taken offline without executive approval. Order the work you can start now.
Answer: Confirm what the trail would have recorded, and its retention window Pull the EDR, firewall and DNS telemetry for the host Capture memory and volatile state Request approval to image the disk
Retention comes first because it is the only clock you do not control. Establishing that proxy logs roll at 30 days changes what you do in the next hour. Pulling the remote telemetry is next: it needs no approval, touches nothing, and cannot be altered by anything on the host. Memory capture affects a production system and is the first step that needs a conversation. Disk imaging needs the approval you are waiting on, so it is last by circumstance rather than by volatility.
Read more: Order of volatility
-
You isolate the host from the network before capturing anything. What did you gain and lose?
Answer: Gained: stopped ongoing damage. Lost: live C2 connections, and you may have tipped off anything watching for connectivity loss
Both effects are real and the trade depends on what is happening. Active encryption or ongoing exfiltration makes isolating first obviously right. A quiet, long-dwell intrusion makes it worth capturing the live channel first. The usual compromise is network isolation that leaves the EDR channel up, which stops the damage while keeping your own visibility. What matters for the report is that you decided, rather than that it happened.
Read more: Order of volatility: collection decisions
-
Your report will say a file was deleted from a temp directory. What has to be recorded at acquisition for that to hold up?
Answer: A hash taken at the point of collection, verified again after transfer
An image without a hash recorded at acquisition is an image whose integrity you cannot demonstrate later. Hash at the point of collection, verify after transfer, and record both. The other three are worth having, and the analyst name and tool version belong in the notes, but none of them establishes that what you are analyzing is what you collected.
Read more: Order of volatility: collection decisions
-
Why does running your own triage commands need recording?
Answer: They create processes, open handles and write artifacts that the next analyst could otherwise attribute to the attacker
Collection is never free. Every command spawns a process, touches access times and can write a prefetch entry. Someone reviewing the host later has no way to tell your netstat from the attacker's unless you wrote down what you ran, when, and as which account. This is also the cheapest thing on the page to get right, and one of the most common things to skip.
Read more: Order of volatility: collection decisions