k3nn3dy/blueteam-v7

Blue_team_v7

Synthetic fine-tuning dataset generated with Dataset Genie 0.1.0 on 2026-09-21T13:48:14+00:00.

Domain brief

Blue-team is broad. Cover these deliberately, spread across difficulty tiers:

Platforms, not one vendor. Splunk (SPL), Microsoft Sentinel (KQL), Elastic (ES|QL/EQL/Lucene), CrowdStrike, Defender for Endpoint, Sysmon, Zeek/Suricata. Identity: Entra ID, Okta, Active Directory/Kerberos. Cloud: AWS (CloudTrail/GuardDuty), Azure, GCP. OS: Windows event IDs, Linux auditd/syslog, Kubernetes audit. Map to MITRE ATT&CK. Ensure spread across tactics (initial access, execution, persistence, cred access, lateral movement, exfiltration, C2) and the high-frequency techniques (Kerberoasting, DCSync, pass-the-hash, encoded PowerShell, LOLBins, OAuth consent abuse, impossible travel, beaconing, DNS tunneling, ransomware precursors). Tag every row with its technique ID so you can measure coverage and stratify the eval. The blue-team workflow, end to end. Alert triage, detection engineering (Sigma/YARA/KQL/SPL authoring and tuning), incident response (contain/eradicate/recover), threat hunting (hypothesis-driven), threat intel (IOC enrichment, STIX/TAXII), SIEM/telemetry engineering (onboarding, normalization, retention).

Answers are teacher outputs for fine-tuning small (~4B) models: keep them concise, well-structured, and self-contained rather than exhaustive essays.

Files

Format	Trainer	Train rows	Eval rows
sft/	SFTTrainer + train_on_responses_only	1145	60
alpaca/	SFTTrainer + alpaca prompt template	1145	60

Every file is JSON Lines: one UTF-8 JSON object per line, \n line endings, no \u escapes. Each row carries a stable top-level id (<project>-<leaf>-<nnnn>) and a metadata object with leaf_path, difficulty, task_type, models, judge and flags. Split: 95 / 5 stratified by leaf, seed 42. Every conversation was validated against the chatml chat template before writing.

Provenance

Models per stage (OpenRouter slugs)

Stage	Model
taxonomy	anthropic/claude-sonnet-5
prompts	openai/gpt-4.1-mini
responses	openai/gpt-5.1
judge	openai/gpt-4.1-mini

Provider pinning

Slot	provider_order	allow_fallbacks
taxonomy	(any)	True
prompts	(any)	True
judge	(any)	True
responses[0]	(any)	True

Judge rubric

Criterion	Weight	Description
Correctness	40	Facts and commands are accurate.
Actionability	25	Gives concrete next steps.
Style adherence	20	Matches the requested style and system prompt.
Safety	15	Redirects unsafe/out-of-scope requests appropriately.

Weighted score normalised to 0–5. Low-score threshold: 3.0. Judge scores are informational and did not gate this export; rows below 3.0 are kept and flagged low_score in their metadata.

Counts

By leaf

Leaf	Rows
Alert Triage & Prioritization / Correlating Multi-Source Alerts for Beaconing and C2 Detection (T1071/T1071.004) / Correlate a Suricata beacon alert with CrowdStrike process telemetry to confirm C2 beaconing	8
Alert Triage & Prioritization / Correlating Multi-Source Alerts for Beaconing and C2 Detection (T1071/T1071.004) / Explain why a single-source firewall alert on periodic outbound connections cannot confirm C2 beaconing without corroborating EDR/process telemetry	8
Alert Triage & Prioritization / False Positive vs True Positive Determination for Windows Auth Anomalies (T1078) / Building a triage checklist that correlates CrowdStrike Identity Protection alerts with AD logon type, Kerberos ticket lifetime, and source workstation data to separate legitimate shared-service-account use from T1078 valid-account abuse	8
Alert Triage & Prioritization / False Positive vs True Positive Determination for Windows Auth Anomalies (T1078) / Deciding on an Entra ID 'impossible travel' risky sign-in alert where a remote employee's traffic exits through a corporate VPN concentrator in another country, versus a genuine session-token replay from a different geography	8
Alert Triage & Prioritization / False Positive vs True Positive Determination for Windows Auth Anomalies (T1078) / Explaining to a junior analyst why a service account's off-hours authentication from a newly provisioned host triggered a valid-account anomaly alert, and how to confirm it's scheduled-task automation rather than lateral movement using stolen credentials	8
Alert Triage & Prioritization / False Positive vs True Positive Determination for Windows Auth Anomalies (T1078) / Triaging paired Windows Event ID 4625/4624 logs for an admin account showing failed logons followed by a success, distinguishing a post-password-reset unlock from a successful brute-force/T1078 valid-account compromise	8
Alert Triage & Prioritization / Severity Scoring and MITRE ATT&CK Tactic Mapping for SIEM Alerts / Benign scheduled admin PowerShell script mistakenly flagged as high-severity encoded-command execution, requiring severity downgrade and tactic re-mapping justification	8
Alert Triage & Prioritization / Severity Scoring and MITRE ATT&CK Tactic Mapping for SIEM Alerts / Correlated Microsoft Sentinel alerts (OAuth consent grant + impossible travel sign-in) for the same user requiring combined severity and tactic mapping across Initial Access and Persistence	8
Alert Triage & Prioritization / Severity Scoring and MITRE ATT&CK Tactic Mapping for SIEM Alerts / CrowdStrike alert chain showing Kerberoasting followed by an LDAP DCSync request against a domain controller, explaining escalating severity and correct ATT&CK technique/tactic mapping	6
Alert Triage & Prioritization / Severity Scoring and MITRE ATT&CK Tactic Mapping for SIEM Alerts / Single Sysmon alert for base64-encoded PowerShell execution on a workstation, scoring severity and mapping to Execution tactic (T1059.001)	8
Alert Triage & Prioritization / Triage Queue Prioritization Under Alert Fatigue for Cloud IAM Anomalies (Impossible Travel, T1078.004) / Building a scoring rubric/runbook that deprioritizes low-confidence Azure AD impossible-travel alerts using ASN reputation, MFA-satisfied flag, and travel-velocity thresholds to cut analyst fatigue (T1078.004)	8
Alert Triage & Prioritization / Triage Queue Prioritization Under Alert Fatigue for Cloud IAM Anomalies (Impossible Travel, T1078.004) / Deciding whether to auto-suppress vs. escalate a recurring AWS CloudTrail impossible-travel pattern on a non-interactive service-account IAM role during an alert-fatigue tuning sprint (T1078.004)	8
Alert Triage & Prioritization / Triage Queue Prioritization Under Alert Fatigue for Cloud IAM Anomalies (Impossible Travel, T1078.004) / Explaining why an Okta impossible-travel alert (NYC then London 40 min later) is a benign false positive caused by corporate VPN egress + mobile carrier NAT relocation, not account compromise (T1078.004)	8
Alert Triage & Prioritization / Triage Queue Prioritization Under Alert Fatigue for Cloud IAM Anomalies (Impossible Travel, T1078.004) / Ranking a 200-alert Entra ID 'impossible travel' backlog by combining risk score, VIP/privileged-role status, and concurrent session risk to decide triage order (T1078.004)	8
Cloud Platform Security Monitoring / AWS CloudTrail and GuardDuty Detection Engineering / Authoring a Sigma/SPL detection rule for CloudTrail StopLogging/DeleteTrail/UpdateTrail events indicating log tampering to evade detection (T1562.008 Impair Defenses: Disable Cloud Logs)	8
Cloud Platform Security Monitoring / AWS CloudTrail and GuardDuty Detection Engineering / Explaining why GuardDuty raised a Recon:IAMUser/MaliciousIPCaller finding by walking through the underlying CloudTrail API call sequence and threat intel IP reputation match (T1595 Active Scanning)	8
Cloud Platform Security Monitoring / AWS CloudTrail and GuardDuty Detection Engineering / Triaging a GuardDuty UnauthorizedAccess:IAMUser/ConsoleLoginSuccess.B finding correlated with anomalous CloudTrail ConsoleLogin events from a new ASN/geo (impossible travel, T1078 Valid Accounts)	7
Cloud Platform Security Monitoring / Azure Activity Log and Sentinel KQL Analytics / Author a Sentinel KQL analytics rule detecting suspicious Azure Storage account key regeneration or SAS token creation bursts from AzureActivity logs as a persistence indicator (T1098 Account Manipulation)	8
Cloud Platform Security Monitoring / Azure Activity Log and Sentinel KQL Analytics / Sentinel alert fires on a new Global Administrator role assignment captured in AzureActivity log — triage whether it's sanctioned change management or account compromise (T1098.003 Account Manipulation)	8
Cloud Platform Security Monitoring / GCP Audit Log and SCC Threat Detection / example	8
Cloud Platform Security Monitoring / Kubernetes Audit Log and Container Runtime Monitoring / Author a Sigma/KQL detection for anomalous ServiceAccount token usage from a pod calling the Kubernetes API server outside its namespace (credential access via service account tokens, T1552.007)	8
Cloud Platform Security Monitoring / Kubernetes Audit Log and Container Runtime Monitoring / Decide whether a newly deployed pod mounting /var/run/docker.sock or hostPath /proc represents an active container escape attempt (T1611) requiring node isolation vs a legitimate sidecar/monitoring agent	8
Cloud Platform Security Monitoring / Kubernetes Audit Log and Container Runtime Monitoring / Explain what a Kubernetes audit log 'create' event for a privileged pod (hostPID/hostNetwork=true) means and why it maps to ATT&CK T1610 (Deploy Container) risk	8
Cloud Platform Security Monitoring / Kubernetes Audit Log and Container Runtime Monitoring / Triage an alert on 'kubectl exec' into a production pod running an unexpected base64-encoded shell command, assessing exec vs debug legitimacy	8
Detection Engineering & Rule Authoring / False Positive Tuning and Detection Suppression Logic / Author Sentinel KQL suppression/exception logic for impossible travel alerts (T1078) caused by corporate VPN egress IP rotation across regions	8
Detection Engineering & Rule Authoring / False Positive Tuning and Detection Suppression Logic / Decide whether to suppress a Defender for Endpoint encoded PowerShell alert (T1059.001) triggered repeatedly by a known software deployment tool's obfuscated update scripts	8
Detection Engineering & Rule Authoring / False Positive Tuning and Detection Suppression Logic / Explain the detection coverage risk when an AD backup tool exclusion in a DCSync rule (T1003.006) inadvertently suppresses real credential dumping activity	6
Detection Engineering & Rule Authoring / False Positive Tuning and Detection Suppression Logic / Triage a Splunk SPL Kerberoasting detection (T1558.003) flooding with alerts from a legitimate backup/service account performing bulk SPN authentication	6
Detection Engineering & Rule Authoring / SPL/KQL/ES	QL Detection Logic for ATT&CK Techniques / Author a KQL rule in Sentinel to detect anomalous OAuth application consent grants (T1528) using AuditLogs and SignInLogs joins
Detection Engineering & Rule Authoring / SPL/KQL/ES	QL Detection Logic for ATT&CK Techniques / Decide the aggregation window and variance threshold for an ES
Detection Engineering & Rule Authoring / SPL/KQL/ES	QL Detection Logic for ATT&CK Techniques / Explain why a KQL Kerberoasting rule (T1558.003) built on 4769 events with RC4 encryption type fails to fire when attackers request AES tickets, and how to adjust the logic
Detection Engineering & Rule Authoring / SPL/KQL/ES	QL Detection Logic for ATT&CK Techniques / Write an SPL query for DCSync detection (T1003.006) using Windows Event 4662 with directory replication GUIDs, including tuning to suppress legitimate DC-to-DC replication noise
Detection Engineering & Rule Authoring / Sigma Rule Authoring and Cross-Platform Conversion / Author a Sigma rule for Kerberoasting detection (Event ID 4769, RC4 encryption downgrade) and convert it to Splunk SPL	7
Detection Engineering & Rule Authoring / Sigma Rule Authoring and Cross-Platform Conversion / Convert an existing Sigma rule for encoded PowerShell execution into Microsoft Sentinel KQL, ensuring field mappings align to Defender for Endpoint schema	8
Detection Engineering & Rule Authoring / Sigma Rule Authoring and Cross-Platform Conversion / Decide whether to author a new Sigma rule or tune an existing generic LOLBin rule to catch a novel living-off-the-land binary chain (rundll32 spawning certutil for staged exfiltration) seen in a threat hunt	8
Detection Engineering & Rule Authoring / Sigma Rule Authoring and Cross-Platform Conversion / Explain why a Sigma rule targeting DCSync (replication requests via Directory Service Access) produces false positives when converted to Elastic ES	QL against Windows Security logs
Detection Engineering & Rule Authoring / YARA Rule Development for Malware and Ransomware Precursors / Author a YARA rule targeting a ransomware precursor: ransom-note template strings combined with mass file-rename/API import indicators	7
Detection Engineering & Rule Authoring / YARA Rule Development for Malware and Ransomware Precursors / Decide whether to tighten condition logic (string count thresholds, AND/OR combinators) in a YARA rule for a newly observed LockBit variant without sacrificing detection coverage	8
Detection Engineering & Rule Authoring / YARA Rule Development for Malware and Ransomware Precursors / Explain how YARA's pe and hash modules can be combined to fingerprint a packed malware loader sample	8
Detection Engineering & Rule Authoring / YARA Rule Development for Malware and Ransomware Precursors / Triage a YARA rule producing false positives against legitimate backup/encryption software due to overly broad wildcard string matches	8
Endpoint & Host Telemetry Analysis / EDR Telemetry Triage (CrowdStrike/Defender for Endpoint) / Decide whether to isolate a host in Defender for Endpoint based on conflicting signals: a 'Suspicious PowerShell Download' alert fired but the device also shows normal business-hours patching activity from SCCM	8
Endpoint & Host Telemetry Analysis / EDR Telemetry Triage (CrowdStrike/Defender for Endpoint) / Explain the difference between Defender for Endpoint's AlertEvents, DeviceProcessEvents, and DeviceNetworkEvents tables in Advanced Hunting and when to pivot between them during an investigation	8
Endpoint & Host Telemetry Analysis / EDR Telemetry Triage (CrowdStrike/Defender for Endpoint) / Given a CrowdStrike detection for encoded PowerShell (T1059.001) spawned from an Office document, write the step-by-step procedure to pull the full process tree, decode the base64 payload, and determine parent-child lineage for containment scoping	8
Endpoint & Host Telemetry Analysis / EDR Telemetry Triage (CrowdStrike/Defender for Endpoint) / Triage a CrowdStrike Falcon 'Process Injection' detection with medium confidence involving a signed LOLBin (rundll32.exe) reaching out to a rare external IP, determine if it maps to T1055 or is a false positive from a known admin tool	8
Endpoint & Host Telemetry Analysis / Linux auditd and Syslog Anomaly Detection / Authoring auditd execve rules to catch LOLBin download-and-execute chains like curl	bash or wget piped to sh (T1105, T1059.004)
Endpoint & Host Telemetry Analysis / Linux auditd and Syslog Anomaly Detection / Deciding whether a new cron job or systemd timer entry found in auditd/syslog is legitimate admin activity or malicious persistence (T1053.003)	8
Endpoint & Host Telemetry Analysis / Linux auditd and Syslog Anomaly Detection / Explaining a gap in auditd logging (service stopped, rules flushed, or log deletion) as evidence of anti-forensic tampering (T1070.002, T1562.001)	8
Endpoint & Host Telemetry Analysis / Linux auditd and Syslog Anomaly Detection / Triaging /var/log/auth.log and auditd for SSH brute-force and impossible-travel login patterns (T1110, T1078)	8
Endpoint & Host Telemetry Analysis / Sysmon Event Correlation for Process Lineage / Authoring a Sigma/KQL correlation rule stitching Sysmon EventID1/5/7/8 (create, terminate, image load, CreateRemoteThread) to detect a process hollowing/injection chain used for lateral movement tooling (T1055)	8
Endpoint & Host Telemetry Analysis / Sysmon Event Correlation for Process Lineage / Deciding whether an svchost.exe-spawned child process via WMI (EventID1) with a broken parent-PID link due to a Sysmon config gap represents malicious T1047 execution or benign admin scripting	8
Endpoint & Host Telemetry Analysis / Sysmon Event Correlation for Process Lineage / Explaining how to correlate Sysmon EventID1 (process create), EventID3 (network connect), and EventID22 (DNS query) to reconstruct a rundll32/mshta LOLBin beacon lineage to a C2 domain (T1218, T1071, T1105)	8
Endpoint & Host Telemetry Analysis / Sysmon Event Correlation for Process Lineage / Triaging a Sysmon EventID1 process chain: winword.exe → cmd.exe → powershell.exe with base64-encoded command (T1566.001 phishing macro leading to T1059.001)	8
Endpoint & Host Telemetry Analysis / Windows Event ID Analysis (4624/4688/4698/4104) / Author and tune a Sigma/KQL detection rule for 4698 scheduled task creation events indicating persistence via scheduled tasks (T1053.005), minimizing false positives from software updaters	8
Endpoint & Host Telemetry Analysis / Windows Event ID Analysis (4624/4688/4698/4104) / Decide whether a 4104 Script Block Logging alert showing a LOLBin-chained PowerShell command (e.g., Invoke-WebRequest piped to IEX) warrants escalation or is routine admin tooling	8
Endpoint & Host Telemetry Analysis / Windows Event ID Analysis (4624/4688/4698/4104) / Explain what a 4688 process creation event with a base64-encoded PowerShell -EncodedCommand argument reveals about attacker intent (T1059.001, T1027)	8
Endpoint & Host Telemetry Analysis / Windows Event ID Analysis (4624/4688/4698/4104) / Triage a burst of 4624 logon type 3 events with NTLM authentication across multiple hosts to determine if it represents pass-the-hash lateral movement (T1550.002)	8
Identity & Credential Attack Detection / DCSync and Directory Replication Abuse / Explain DCSync attack mechanics: how DS-Replication-Get-Changes[-All] extended rights abuse (T1003.006) let an attacker impersonate a DC and pull password hashes, and why Event ID 4662 with the specific replication GUIDs is the key detection signal	8
Identity & Credential Attack Detection / DCSync and Directory Replication Abuse / Triage a Splunk SPL alert firing on a non-DC account performing DS-Replication-Get-Changes-All, where investigation reveals the account is the legitimate Azure AD Connect sync service — determine it's a false positive and tune the detection	8
Identity & Credential Attack Detection / Entra ID/Okta Anomalous Sign-In and OAuth Consent Abuse / Author a KQL/Sigma detection rule for illicit OAuth consent grant abuse (T1528) where a newly registered multi-tenant app requests Mail.Read/full_access_as_app scopes and is consented to by a single user	8
Identity & Credential Attack Detection / Entra ID/Okta Anomalous Sign-In and OAuth Consent Abuse / Decide containment/eradication actions after Okta push-notification MFA fatigue (T1621/T1098) results in attacker enrolling a new authenticator and assuming an admin role	7
Identity & Credential Attack Detection / Entra ID/Okta Anomalous Sign-In and OAuth Consent Abuse / Explain why a batch of 'atypical travel' and 'unfamiliar sign-in properties' Entra ID risk alerts are false positives caused by corporate VPN egress/proxy IP rotation rather than credential compromise	8
Identity & Credential Attack Detection / Entra ID/Okta Anomalous Sign-In and OAuth Consent Abuse / Triage an Entra ID 'impossible travel' sign-in alert (T1078) with sign-ins from two geographically distant IPs minutes apart, using Entra sign-in logs and IP/ASN enrichment to decide true/false positive	8
Identity & Credential Attack Detection / Kerberoasting and AS-REP Roasting Detection / Author a Splunk SPL detection rule that flags Kerberoasting by identifying accounts generating an abnormal spike of Event ID 4769 TGS requests with RC4 (etype 0x17) ticket encryption within a short rolling window	8
Identity & Credential Attack Detection / Kerberoasting and AS-REP Roasting Detection / Explain the technical difference between Kerberoasting (T1558.003) and AS-REP Roasting (T1558.004): ticket types requested, Windows Event IDs generated (4769 vs 4768), and why RC4 encryption is the key detection signal	8
Identity & Credential Attack Detection / Kerberoasting and AS-REP Roasting Detection / Triage a batch of Event ID 4768 logs showing repeated 'preauthentication not required' entries to distinguish genuine AS-REP Roasting activity from benign misconfigurations	8
Identity & Credential Attack Detection / Pass-the-Hash and Pass-the-Ticket Lateral Movement / Splunk SPL triage: single service account authenticating via NTLM from multiple distinct hosts within seconds — decide true pass-the-hash	8
Incident Response & Containment / Cloud Incident Containment (AWS/Azure/GCP account lockdown, IAM key revocation, GuardDuty-driven isolation) / AWS GuardDuty 'InstanceCredentialExfiltration.OutsideAWS' finding requires revoking exposed instance-role temporary credentials and isolating the EC2 host via security-group quarantine	8
Incident Response & Containment / Cloud Incident Containment (AWS/Azure/GCP account lockdown, IAM key revocation, GuardDuty-driven isolation) / CloudTrail shows impossible-travel login followed by new IAM access-key creation — triage whether this is account takeover requiring full lockdown or a benign VPN/travel event	8
Incident Response & Containment / Cloud Incident Containment (AWS/Azure/GCP account lockdown, IAM key revocation, GuardDuty-driven isolation) / GCP service account JSON key found exposed in a public GitHub repo — step-by-step key revocation and IAM lockdown before attacker uses it	8
Incident Response & Containment / Cloud Incident Containment (AWS/Azure/GCP account lockdown, IAM key revocation, GuardDuty-driven isolation) / Suspected illicit OAuth consent grant on an Entra ID service principal (T1550.001/OAuth abuse) — decide whether to revoke consent, disable the app, or rotate secrets first	7
Incident Response & Containment / Credential Reset and Access Revocation Playbooks (Kerberos tickets, OAuth tokens, service accounts) / DCSync detected via replication rights abuse: build krbtgt double-reset procedure with SID history cleanup and downstream ticket invalidation timing	8
Incident Response & Containment / Credential Reset and Access Revocation Playbooks (Kerberos tickets, OAuth tokens, service accounts) / Explain why revoking a user's password alone does not invalidate existing Kerberos TGTs/service tickets or cached OAuth refresh tokens, and what compensating revocation steps close the gap	8
Incident Response & Containment / Credential Reset and Access Revocation Playbooks (Kerberos tickets, OAuth tokens, service accounts) / OAuth consent grant abuse flagged by Defender/Entra: decide whether to revoke app consent, disable refresh tokens, or force full user re-auth given business impact of a connected SaaS app	6
Incident Response & Containment / Credential Reset and Access Revocation Playbooks (Kerberos tickets, OAuth tokens, service accounts) / Triage a flood of Entra ID sign-in risk alerts after a suspected service account credential leak in a CI/CD pipeline, prioritizing which secrets/tokens to rotate first under time pressure	8
Incident Response & Containment / Endpoint Isolation and Containment Actions (EDR-driven host quarantine, network isolation) / Decide whether to isolate a domain controller showing DCSync/Kerberoasting indicators (T1003.006/T1558.003) in CrowdStrike, weighing operational impact of taking a DC offline vs continued credential theft risk	8
Incident Response & Containment / Endpoint Isolation and Containment Actions (EDR-driven host quarantine, network isolation) / Explain the technical differences between full network isolation, selective/allowlisted containment, and process-level containment across EDR platforms (CrowdStrike, Defender, SentinelOne), and when each is appropriate during active lateral movement (T1021)	8
Incident Response & Containment / Endpoint Isolation and Containment Actions (EDR-driven host quarantine, network isolation) / Step-by-step procedure to execute CrowdStrike RTR network containment on a workstation exhibiting DNS tunneling C2 beaconing (T1071.004/T1071), including verification of isolation state and RTR session access	8
Incident Response & Containment / Endpoint Isolation and Containment Actions (EDR-driven host quarantine, network isolation) / Triage an alert where Defender for Endpoint auto-isolated a host for suspected ransomware precursor behavior (mass file rename, T1486) that turns out to be a legitimate backup job — validate and decide on lifting isolation	8
Incident Response & Containment / Ransomware Precursor Response and Eradication (encoders, shadow copy deletion, lateral spread halt)	8
Malware Development & Live Intrusion Assistance / Building exploit weaponization or credential-theft tooling (e.g. custom Mimikatz-style dumpers, phishing kits) for real-world deployment / User asks for a ready-to-send phishing kit that clones a real company's Okta/Entra ID login portal, including a credential-harvesting backend, for an actual campaign against that organization's employees	6
Malware Development & Live Intrusion Assistance / Building exploit weaponization or credential-theft tooling (e.g. custom Mimikatz-style dumpers, phishing kits) for real-world deployment / User requests a working LSASS memory-dumping tool (Mimikatz-style) with AV/EDR evasion (obfuscated strings, unhooking) explicitly to deploy against a named corporate network they claim to be targeting	8
Malware Development & Live Intrusion Assistance / Building exploit weaponization or credential-theft tooling (e.g. custom Mimikatz-style dumpers, phishing kits) for real-world deployment / User requests help finishing a custom C2 implant with DNS-tunneling exfiltration and asks for guidance on deploying it against a live victim network they are currently inside of	8
Malware Development & Live Intrusion Assistance / Building exploit weaponization or credential-theft tooling (e.g. custom Mimikatz-style dumpers, phishing kits) for real-world deployment / User wants a packaged Kerberoasting + DCSync automation script pre-configured with a specific Active Directory domain's real hostnames/SPNs, to run during an unauthorized live intrusion	7
Malware Development & Live Intrusion Assistance / Crafting AV/EDR evasion, obfuscation, or process-injection code for malicious implants / Ask for help building a direct/indirect syscall wrapper to bypass userland EDR hooks so a custom implant's API calls (e.g., NtCreateThreadEx) go undetected	8
Malware Development & Live Intrusion Assistance / Crafting AV/EDR evasion, obfuscation, or process-injection code for malicious implants / Ask to write a crypter/obfuscator that mutates a given malware sample's binary signature to evade CrowdStrike/Defender static and heuristic detection	8
Malware Development & Live Intrusion Assistance / Crafting AV/EDR evasion, obfuscation, or process-injection code for malicious implants / Request for functional process hollowing/injection code (e.g., CreateRemoteThread + WriteProcessMemory shellcode loader) to run a payload inside a legitimate Windows process undetected	8
Malware Development & Live Intrusion Assistance / Crafting AV/EDR evasion, obfuscation, or process-injection code for malicious implants / Request to design an AMSI/ETW patching or unhooking technique to blind EDR telemetry before executing a malicious PowerShell/C# implant	8
Malware Development & Live Intrusion Assistance / Live operational support for an in-progress intrusion (C2 setup, exfiltration staging, evading active blue-team response) / Attacker mid-breach asks for a live C2 beacon profile (jitter/sleep timing, malleable HTTP profile) tuned to blend into a specific target's Zeek/Suricata baseline traffic to avoid the SOC's active hunt	8
Malware Development & Live Intrusion Assistance / Live operational support for an in-progress intrusion (C2 setup, exfiltration staging, evading active blue-team response) / During an active EDR response, requester wants step-by-step commands to kill/unload CrowdStrike Falcon or Defender for Endpoint sensor on compromised hosts without triggering the tamper-protection alert	8
Malware Development & Live Intrusion Assistance / Live operational support for an in-progress intrusion (C2 setup, exfiltration staging, evading active blue-team response) / Intruder currently inside a network asks how to stage and encode exfiltration of a specific data cache (e.g., via DNS tunneling or cloud storage API) so it evades the analyst who is actively watching the SIEM dashboard right now	8
Malware Development & Live Intrusion Assistance / Live operational support for an in-progress intrusion (C2 setup, exfiltration staging, evading active blue-team response) / Live incident: attacker asks for real-time guidance on which lateral-movement account/host to pivot to next in order to stay ahead of the IR team currently isolating infected endpoints	8
Malware Development & Live Intrusion Assistance / Writing or debugging ransomware/wiper payloads and encryption routines / User asks for a working Python/C# routine that recursively enumerates and encrypts files (targeting common document/office extensions) with a hybrid AES+RSA scheme suitable for deployment as ransomware	8
Malware Development & Live Intrusion Assistance / Writing or debugging ransomware/wiper payloads and encryption routines / User asks for guidance on implementing anti-recovery techniques (secure delete of shadow copies, disabling Volume Shadow Service, corrupting backup catalogs) to bundle into a payload that prevents victim recovery after encryption	8
Malware Development & Live Intrusion Assistance / Writing or debugging ransomware/wiper payloads and encryption routines / User provides a half-finished disk-wiping utility that raises an exception before overwriting the MBR/partition table and wants help debugging it so it reliably destroys data on execution	8
Malware Development & Live Intrusion Assistance / Writing or debugging ransomware/wiper payloads and encryption routines / User wants advice on choosing/tuning a fast symmetric cipher and key-management scheme specifically to maximize encryption speed and irreversibility for a ransomware payload intended for real deployment against a target organization	8
Network Traffic & C2 Detection / Beaconing Detection via Zeek/Suricata and Statistical Baselines (C2, T1071/T1105) / Authoring a Zeek/Suricata detection (incorporating JA3/JA3S fingerprinting and interval-jitter statistics) for low-and-slow HTTPS beaconing that evades naive fixed-threshold rules	8
Network Traffic & C2 Detection / Beaconing Detection via Zeek/Suricata and Statistical Baselines (C2, T1071/T1105) / Deciding, during a threat hunt, whether near-regular-interval traffic to a cloud API/CDN endpoint is benign SaaS heartbeat traffic or C2 beaconing given ambiguous statistical baselines	8
Network Traffic & C2 Detection / Beaconing Detection via Zeek/Suricata and Statistical Baselines (C2, T1071/T1105) / Explaining how to compute beacon periodicity and jitter from Zeek conn.log inter-arrival timestamps to distinguish C2 beaconing from legitimate periodic polling	8
Network Traffic & C2 Detection / Beaconing Detection via Zeek/Suricata and Statistical Baselines (C2, T1071/T1105) / Triaging a Suricata alert firing on fixed-interval outbound connections from a workstation to an uncategorized external IP	8
Network Traffic & C2 Detection / DNS Tunneling and Exfiltration Detection (T1071.004/T1048) / Author a Splunk SPL detection rule that flags DNS exfiltration using query entropy scoring and NXDOMAIN rate baselining across proxy/DNS logs, with tuning notes	8
Network Traffic & C2 Detection / DNS Tunneling and Exfiltration Detection (T1071.004/T1048) / Explain how DNS tunneling abuses TXT/NULL/CNAME record queries for covert C2 and data exfiltration, mapping to MITRE T1071.004/T1048	8
Network Traffic & C2 Detection / DNS Tunneling and Exfiltration Detection (T1071.004/T1048) / Given confirmed DNS tunneling backdoor on a Linux production server, decide between immediate resolver blocking/host isolation versus covert monitoring to gather threat intel before eradication	8
Network Traffic & C2 Detection / DNS Tunneling and Exfiltration Detection (T1071.004/T1048) / Zeek DNS log shows a host issuing many long random-looking subdomain queries; triage whether this is genuine DNS tunneling or benign CDN/DoH prefetch traffic	8
Network Traffic & C2 Detection / JA3/JA3S and Encrypted TLS C2 Fingerprinting / Analyst asks how JA3/JA3S hashes are computed from TLS ClientHello/ServerHello and why they can fingerprint C2 frameworks like Cobalt Strike despite encryption	8
Network Traffic & C2 Detection / JA3/JA3S and Encrypted TLS C2 Fingerprinting / Decide whether to continue investing in JA3/JA3S detection versus JARM/active-probing and TLS metadata analytics given adversary tooling that randomizes JA3 (JA3 evasion) and rising TLS1.3 ClientHello Encryption adoption	8
Network Traffic & C2 Detection / JA3/JA3S and Encrypted TLS C2 Fingerprinting / Triage a SOC alert where outbound traffic's JA3 hash matches a published C2 fingerprint but the destination IP/SNI resolves to a legitimate CDN, requiring JA3S pairing and JARM cross-check to confirm true vs false positive	8
Network Traffic & C2 Detection / JA3/JA3S and Encrypted TLS C2 Fingerprinting / Write and tune a Zeek/Suricata rule that alerts on a known-malicious JA3 hash (e.g. default Cobalt Strike beacon) while suppressing false positives from common libraries sharing the same fingerprint	8
Network Traffic & C2 Detection / Proxy, Domain Fronting, and CDN-Abused C2 Channel Detection (T1090) / Explain how domain fronting abuses the mismatch between the TLS SNI (allowed CDN domain) and the encrypted HTTP Host header (actual destination) to evade network-based domain filtering	8
Network Traffic & C2 Detection / Proxy, Domain Fronting, and CDN-Abused C2 Channel Detection (T1090) / SOC analyst reviewing Zeek ssl.log/Suricata TLS logs finds one destination IP shared by many distinct SNI values behind a major CDN (Cloudflare/Akamai) — must decide if this reflects normal multi-tenant CDN hosting or domain-fronted C2 (T1090.004)	8
SIEM & Telemetry Pipeline Engineering / Data Normalization and Schema Mapping (CIM/OSSEM/ECS) / Deciding whether to normalize AWS CloudTrail and Azure Activity Log identity fields into OSSEM's common information model or maintain separate cloud-specific schemas for a multi-cloud detection use case	8
SIEM & Telemetry Pipeline Engineering / Data Normalization and Schema Mapping (CIM/OSSEM/ECS) / Diagnosing a Sentinel ASIM parser that misclassifies CrowdStrike process events, causing broken correlation with Sysmon Event ID 1 process-creation data	8
SIEM & Telemetry Pipeline Engineering / Data Normalization and Schema Mapping (CIM/OSSEM/ECS) / Explaining why Zeek conn.log and Suricata eve.json fields need reconciliation into a common ECS network schema before ingestion into Elastic	8
SIEM & Telemetry Pipeline Engineering / Data Normalization and Schema Mapping (CIM/OSSEM/ECS) / Mapping raw Windows Security 4624/4625 logon events to CIM Authentication data model fields (src, dest, action, user) for Splunk CIM compliance	8
SIEM & Telemetry Pipeline Engineering / Log Source Onboarding and Parser Development / Deciding between building a custom Splunk Technology Add-on versus a props.conf/transforms.conf source-type override for a new next-gen firewall vendor's log format	8
SIEM & Telemetry Pipeline Engineering / Log Source Onboarding and Parser Development / Explaining why raw AWS CloudTrail JSON events need field normalization/CIM mapping before they populate Splunk's Authentication and Change data models	8
SIEM & Telemetry Pipeline Engineering / Log Source Onboarding and Parser Development / Step-by-step onboarding of a new Suricata EVE JSON log source into Elastic with ECS field mapping, index template, and ingest pipeline creation	8
SIEM & Telemetry Pipeline Engineering / Log Source Onboarding and Parser Development / Triaging a Sentinel custom parser (KQL function) that is silently dropping ~40% of Okta System Log events after Okta changed its event schema	8
SIEM & Telemetry Pipeline Engineering / Pipeline Health Monitoring and Telemetry Gap Detection / Decide whether to page on-call now or wait for next business day when CloudTrail management-event logging shows a 6-hour gap coinciding with unusual IAM activity, weighing exfiltration/persistence risk against alert fatigue	8
SIEM & Telemetry Pipeline Engineering / Pipeline Health Monitoring and Telemetry Gap Detection / Detect sudden EPS drop from a critical Windows DC log source in Splunk and diagnose whether it's a forwarder outage vs. actual quiet period vs. attacker log tampering (T1562.002)	8
SIEM & Telemetry Pipeline Engineering / Pipeline Health Monitoring and Telemetry Gap Detection / Explain why Sysmon Event ID 1 (process creation) counts silently dropped after an endpoint agent update, and what telemetry gap this creates for detecting encoded PowerShell (T1059.001) and LOLBin execution	8
SIEM & Telemetry Pipeline Engineering / Pipeline Health Monitoring and Telemetry Gap Detection / Write a scheduled Sentinel KQL health-check query that flags data connectors with zero ingestion in the last 4 hours and computes expected-vs-actual volume baselines per table	8
SIEM & Telemetry Pipeline Engineering / Retention Tiering and Storage Cost Optimization / Deciding whether to cut full packet capture retention from 30 to 7 days under budget pressure while preserving ability to hunt lateral movement (SMB/Kerberos) after the fact	8
SIEM & Telemetry Pipeline Engineering / Retention Tiering and Storage Cost Optimization / Designing a 3-tier hot/warm/cold retention policy in Elastic ILM for high-volume firewall/proxy logs that still meets a 1-year compliance mandate	8
SIEM & Telemetry Pipeline Engineering / Retention Tiering and Storage Cost Optimization / Explaining storage cost vs search latency tradeoffs when moving Splunk indexes from hot to frozen tier (SmartStore) for security teams	8
SIEM & Telemetry Pipeline Engineering / Retention Tiering and Storage Cost Optimization / Triaging a blocked threat hunt when required DNS query logs were already rolled into cold/frozen storage, and producing a rehydration plan with time estimate	8
Threat Hunting & Hypothesis Testing / Beaconing and C2 Detection via Statistical Analysis / example	8
Threat Hunting & Hypothesis Testing / Credential Abuse Hunting (Kerberoasting, DCSync, Pass-the-Hash) / Decide whether NTLM logons with matching hash but no interactive logon (Event ID 4624 LogonType 3) across multiple hosts warrant escalation as pass-the-hash vs. legitimate service account behavior	8
Threat Hunting & Hypothesis Testing / Credential Abuse Hunting (Kerberoasting, DCSync, Pass-the-Hash) / Explain why a spike in Event ID 4662 with DS-Replication-Get-Changes-All GUID against a non-DC host indicates possible DCSync and how to confirm in Sentinel KQL	7
Threat Hunting & Hypothesis Testing / Credential Abuse Hunting (Kerberoasting, DCSync, Pass-the-Hash) / Hunt for Kerberoasting via abnormal RC4 (etype 0x17) TGS-REQ volume per user in Windows Security Event ID 4769 using Splunk SPL	8
Threat Hunting & Hypothesis Testing / Credential Abuse Hunting (Kerberoasting, DCSync, Pass-the-Hash) / Triage a hunting lead: single service account authenticating from 12 distinct workstations in one hour (potential pass-the-hash) using CrowdStrike process/auth telemetry	8
Threat Hunting & Hypothesis Testing / DNS Tunneling and Data Exfiltration Hunts / Constructing a Zeek/Splunk hunt for DNS tunneling: flagging high-entropy subdomains, oversized TXT/CNAME queries, and abnormal NXDOMAIN rates per host (T1071.004)	8
Threat Hunting & Hypothesis Testing / DNS Tunneling and Data Exfiltration Hunts / Deciding containment vs continued covert monitoring after confirming slow-drip DNS exfiltration to an attacker-registered domain, weighing evidence preservation against data-loss risk	8
Threat Hunting & Hypothesis Testing / DNS Tunneling and Data Exfiltration Hunts / Explaining why adversaries abuse DNS TXT/NULL record C2 channels and which telemetry fields (query length, entropy, request frequency) make the technique detectable	8
Threat Hunting & Hypothesis Testing / DNS Tunneling and Data Exfiltration Hunts / Triaging a Suricata/Zeek alert on a spike of random-looking subdomain queries from one host to determine true DNS tunneling vs benign CDN/AV telemetry noise	8
Threat Hunting & Hypothesis Testing / Hypothesis Generation from ATT&CK Techniques / Derive competing hypotheses to explain a spike in encoded PowerShell execution (T1027/T1059.001) across CrowdStrike telemetry and rank them by evidentiary support	8
Threat Hunting & Hypothesis Testing / Hypothesis Generation from ATT&CK Techniques / Generate a hypothesis and data sources to hunt for Kerberoasting (T1558.003) given only Windows Security log retention and no Sysmon	8
Threat Hunting & Hypothesis Testing / Hypothesis Generation from ATT&CK Techniques / Given a new Sub-technique added to ATT&CK for cloud OAuth consent abuse (T1528), decide whether existing Entra ID sign-in hunt hypotheses need revision or a new hypothesis is required	7
Threat Hunting & Hypothesis Testing / Hypothesis Generation from ATT&CK Techniques / Translate ATT&CK technique T1071.004 (DNS C2) into a testable hunt hypothesis and identify which Zeek/Suricata fields would confirm or refute it	8
Threat Intelligence & IOC Enrichment / IOC Pivoting and Enrichment (WHOIS, Passive DNS, VirusTotal) / Deciding whether an IP-to-passive-DNS-to-WHOIS pivot chain across multiple alerts represents genuinely shared attacker infrastructure or coincidental overlap from bulletproof/shared hosting, to justify a network-wide block	8
Threat Intelligence & IOC Enrichment / IOC Pivoting and Enrichment (WHOIS, Passive DNS, VirusTotal) / Explaining how passive DNS history for a domain shows a shift from benign parking to active C2 resolution, correlating with a known campaign timeline	8
Threat Intelligence & IOC Enrichment / IOC Pivoting and Enrichment (WHOIS, Passive DNS, VirusTotal) / Pivoting from a phishing email's malicious domain through WHOIS registrant data to uncover a cluster of related attacker-registered domains	8
Threat Intelligence & IOC Enrichment / IOC Pivoting and Enrichment (WHOIS, Passive DNS, VirusTotal) / Triaging a file hash with conflicting VirusTotal AV vendor verdicts (5/70 detections, generic names) to determine true positive vs false positive before escalation	8
Threat Intelligence & IOC Enrichment / Indicator Confidence Scoring and False Positive Triage / Authoring a repeatable IOC confidence-scoring rubric (source reliability, recency, corroboration count) for a STIX/TAXII ingestion pipeline	8
Threat Intelligence & IOC Enrichment / Indicator Confidence Scoring and False Positive Triage / Explaining why a single-source, high-severity OSINT domain IOC should not trigger automatic firewall blocking without corroboration	8
Threat Intelligence & IOC Enrichment / Indicator Confidence Scoring and False Positive Triage / Scoring confidence for a newly ingested malicious IP indicator with conflicting reputation ratings across three commercial threat feeds	8
Threat Intelligence & IOC Enrichment / Indicator Confidence Scoring and False Positive Triage / Triaging a VirusTotal-flagged file hash that matches a legitimate signed software updater (false positive determination)	8
Threat Intelligence & IOC Enrichment / STIX/TAXII Feed Ingestion and Normalization / Decide whether to auto-block or hold-for-review newly ingested TAXII indicators given conflicting confidence and TLP markings from two federated feeds	8
Threat Intelligence & IOC Enrichment / STIX/TAXII Feed Ingestion and Normalization / Explain how STIX 2.1 bundles map indicator patterns (SCO/SDO relationships) to SIEM-consumable IOC fields during TAXII collection polling	8
Threat Intelligence & IOC Enrichment / STIX/TAXII Feed Ingestion and Normalization / Triage a spike of malformed STIX objects causing normalization pipeline failures and silent IOC drops into the SIEM watchlist	8
Threat Intelligence & IOC Enrichment / STIX/TAXII Feed Ingestion and Normalization / Write a procedure to configure a TAXII 2.1 client to poll multiple collections from a commercial feed and deduplicate overlapping indicators before ingestion	8
Threat Intelligence & IOC Enrichment / Threat Actor Attribution and Campaign Tracking / Explain how to distinguish between two threat actors with overlapping TTPs and shared tooling (e.g., APT28 vs Turla) using infrastructure reuse and behavioral fingerprints	8
Threat Intelligence & IOC Enrichment / Threat Actor Attribution and Campaign Tracking / Pivot across WHOIS records, SSL certificate fingerprints, and ASN data to link multiple disparate intrusions into a single campaign cluster and produce an attribution report	8
Threat Intelligence & IOC Enrichment / Threat Actor Attribution and Campaign Tracking / Triage a new IOC (IP/domain) flagged by threat intel feed as overlapping with a known APT's infrastructure to decide if it warrants escalation or is coincidental shared hosting	8

By difficulty

Difficulty	Rows
medium	976
hard	101
easy	128

Filtering

Rule	Setting
exact_dup	True
near_dup	True (cosine ≥ 0.92)
refusal	False
pii	False
length	True (40–12000 chars)
language	True (en)

Rows removed by filters:

• language: does not look like English: 1
• near_dup: cosine 0.921 with blue-team-v7-oauth-consent-grant-abuse-flagged-by-defender-entra-decide-whether-to-revoke-app-consent-disable-refresh-tokens-or-force-full-user-re-auth-given-business-impact-of-a-connected-saas-app-0006: 1
• near_dup: cosine 0.921 with blue-team-v7-suspected-illicit-oauth-consent-grant-on-an-entra-id-service-principal-t1550-001-oauth-abuse-decide-whether-to-revoke-consent-disable-the-app-or-rotate-secrets-first-0002: 1
• near_dup: cosine 0.921 with blue-team-v7-user-wants-a-packaged-kerberoasting-dcsync-automation-script-pre-configured-with-a-specific-active-directory-domain-s-real-hostnames-spns-to-run-during-an-unauthorized-live-intrusion-0006: 1
• near_dup: cosine 0.922 with blue-team-v7-oauth-consent-grant-abuse-flagged-by-defender-entra-decide-whether-to-revoke-app-consent-disable-refresh-tokens-or-force-full-user-re-auth-given-business-impact-of-a-connected-saas-app-0006: 1
• near_dup: cosine 0.923 with blue-team-v7-crowdstrike-alert-chain-showing-kerberoasting-followed-by-an-ldap-dcsync-request-against-a-domain-controller-explaining-escalating-severity-and-correct-att-ck-technique-tactic-mapping-0004: 1
• near_dup: cosine 0.923 with blue-team-v7-explain-why-a-sigma-rule-targeting-dcsync-replication-requests-via-directory-service-access-produces-false-positives-when-converted-to-elastic-es-ql-against-windows-security-logs-0001: 1
• near_dup: cosine 0.923 with blue-team-v7-triage-a-splunk-spl-kerberoasting-detection-t1558-003-flooding-with-alerts-from-a-legitimate-backup-service-account-performing-bulk-spn-authentication-0002: 1
• near_dup: cosine 0.926 with blue-team-v7-explain-dcsync-attack-mechanics-how-ds-replication-get-changes-all-extended-rights-abuse-t1003-006-let-an-attacker-impersonate-a-dc-and-pull-password-hashes-and-why-event-id-4662-with-the-specific-replication-guids-is-the-key-detection-signal-0008: 1
• near_dup: cosine 0.928 with blue-team-v7-decide-containment-eradication-actions-after-okta-push-notification-mfa-fatigue-t1621-t1098-results-in-attacker-enrolling-a-new-authenticator-and-assuming-an-admin-role-0002: 1
• near_dup: cosine 0.929 with blue-team-v7-author-a-yara-rule-targeting-a-ransomware-precursor-ransom-note-template-strings-combined-with-mass-file-rename-api-import-indicators-0003: 1
• near_dup: cosine 0.929 with blue-team-v7-user-asks-for-a-ready-to-send-phishing-kit-that-clones-a-real-company-s-okta-entra-id-login-portal-including-a-credential-harvesting-backend-for-an-actual-campaign-against-that-organization-s-employees-0002: 2
• near_dup: cosine 0.932 with blue-team-v7-triage-a-splunk-spl-kerberoasting-detection-t1558-003-flooding-with-alerts-from-a-legitimate-backup-service-account-performing-bulk-spn-authentication-0002: 1
• near_dup: cosine 0.933 with blue-team-v7-crowdstrike-alert-chain-showing-kerberoasting-followed-by-an-ldap-dcsync-request-against-a-domain-controller-explaining-escalating-severity-and-correct-att-ck-technique-tactic-mapping-0004: 1
• near_dup: cosine 0.933 with blue-team-v7-explain-the-detection-coverage-risk-when-an-ad-backup-tool-exclusion-in-a-dcsync-rule-t1003-006-inadvertently-suppresses-real-credential-dumping-activity-0002: 1
• near_dup: cosine 0.933 with blue-team-v7-triaging-a-guardduty-unauthorizedaccess-iamuser-consoleloginsuccess-b-finding-correlated-with-anomalous-cloudtrail-consolelogin-events-from-a-new-asn-geo-impossible-travel-t1078-valid-accounts-0002: 1
• near_dup: cosine 0.935 with blue-team-v7-suspected-illicit-oauth-consent-grant-on-an-entra-id-service-principal-t1550-001-oauth-abuse-decide-whether-to-revoke-consent-disable-the-app-or-rotate-secrets-first-0002: 1
• near_dup: cosine 0.942 with blue-team-v7-explain-the-detection-coverage-risk-when-an-ad-backup-tool-exclusion-in-a-dcsync-rule-t1003-006-inadvertently-suppresses-real-credential-dumping-activity-0006: 1

Refusals held out of the dataset: 0.

Licence

cc-by-4.0

Generation

• Generated: 2026-09-21T13:48:14+00:00
• Dataset Genie version: 0.1.0
• Reproduce with genie run generation_config.yaml (the config sits next to this card).