Design Partner Program
DETECTION VALIDATION PLATFORM

Prove your detections fire on real attack techniques.

A green rule in your SIEM only means it ran. Hayrok executes approved techniques in your environment and records what each layer blocked, logged, alerted, and routed, so you see exactly where a signal stopped.

run_id_0914Detection run · 6 techniques · web tier → cloud storage
RUN COMPLETE ILLUSTRATIVE DATA
TECHNIQUES RUN6 / 6
DETECTED3
PARTIAL1
MISSED2
TECHNIQUE PREVENT LOG ALERT ROUTE RESULT
00:18 / 00:18
T1190
T1059.004
T1552.005
T1078.004
T1530
T1567.002
T1567.002 · STEP 6 / 6Detected
Upload blocked and the on-call was paged. The proxy blocked the upload of synthetic data and the alert routed as P1.
executedupload synthetic.zip
expectedblock · alert · P1
observedblocked · P1 paged
time to alert2m 01s
TELEMETRY SOURCESSplunkElasticCrowdStrikeDatadogCloudflareAkamai
01 · DETECTION VALIDATION COVERAGE

Detection Coverage: Every Layer That Should Have Seen It

A detection is a chain, not a rule. Each run checks the controls that should block, the sources that should log, the rules that should fire, and the queue the alert should land in.

01
SIEM

Detection rules

Whether each rule fires on the technique it was written for, including the variants attackers actually use.

→rule fires on real technique→variant and evasion coverage
02
EDR AND XDR

Endpoint detections

Process, file, and network behaviour on hosts, and whether the agent raised and forwarded the alert.

→process lineage alerts→agent health and forwarding
03
WAF AND API GATEWAY

Edge prevention

Which payload classes and encodings are blocked, logged only, or passed straight to the application.

→block vs log-only mode→encoding and mutation bypass
04
CLOUD AUDIT LOGS

Cloud control plane

Role assumption, storage access, and logging changes, from audit event to parsed field to rule.

→audit event collected→field mapping intact
05
IDENTITY SIGNALS

Sign-in and privilege

Anomalous sign-ins, privilege changes, and impersonation, checked against identity provider and SIEM rules.

→privilege change alerts→impersonation detection
06
ALERT ROUTING

Who actually sees it

Severity, queue, and playbook handling, so an alert that fires is also an alert someone reads.

→severity and queue→auto-close suppression
02 · HOW DETECTION VALIDATION WORKS

How Detection Validation Works: From Plan to Proof

Connect the tools that should see an attack, pick the techniques they should catch, run them safely, and read what actually happened at every stage.

connections.yamlILLUSTRATIVE
1sources:
2 siem: search API · read
3 edr: alert feed · read
4 waf: log stream · read
5 cloud_audit: org trail · read
6 routing: SOAR cases · read
7
8status: 5 of 5 sources reporting
03 · SIEM DETECTION RULE TESTING

SIEM Detection Testing: Find the Stage That Went Quiet

A rule can be enabled, tested, and still never fire. Hayrok follows each technique through collection, parsing, rule logic, alerting, and routing, and records the first stage where the signal stopped.

T1078.004 · Cloud accountsrule: unusual_role_assumption ILLUSTRATIVE
✓ Executedtechnique ran
✓ Collectedevent in SIEM
! Parsedsignal stopped here
4 Rule matchednot reached
5 Alert raisednot reached
6 Routednot reached
FAILED AT · PARSEDMissed

Logged, but the field the rule needs arrived empty.

A log source update changed the event schema. The rule still runs, and still shows healthy, but matches on a field that is no longer populated.
NEXTFix the field mapping, then replay the technique to confirm the rule fires.
EVIDENCE · EXPECTED VS OBSERVED
eventAssumeRole · collected
rule fielduser.arn
observed valuenull
sinceparser update · 2026-10-14
04 · WAF EFFECTIVENESS TESTING

WAF Validation: What Happens Once It’s Encoded?

Attackers don’t send the textbook payload. Hayrok replays each payload class through the encodings and mutations attackers use, against your WAF in block mode, and records which ones get through.

ENCODING
PAYLOAD CLASSTEST CASEEXPECTEDOBSERVEDRESULT
SQL injection sqli.boolean_tautology · double url block 403 · blocked Blocked
Cross-site scripting xss.event_handler · double url block 200 · passed Passed
Path traversal traversal.dot_segments · double url block 200 · passed Passed
Command injection cmdi.chained_operator · double url block 403 · blocked Blocked
Server-side request forgery ssrf.internal_address · double url block 200 · logged only Logged only
JNDI lookup jndi.lookup_in_header · double url block 403 · blocked Blocked
WAF MODE · BLOCK
3/ 6 blocked
Double-encoded XSS and traversal passed. The WAF decodes once before matching. Test cases use benign markers, not working exploits. A pass means the request reached the application unfiltered.
05 · MITRE ATT&CK DETECTION COVERAGE

MITRE ATT&CK Coverage: Measured, Not Mapped

Mapping rules to techniques shows what you intended to detect. Each cell here is the result of running the technique and checking what fired.

Initial access
Execution
Persistence
Privilege escalation
Defense evasion
Credential access
Discovery
Lateral movement
Collection
Exfiltration
T1552.005Cloud instance metadata API MissedRan in the last cycle. Telemetry may exist, but no expected alert fired. MITRE ATT&CK ↗
06 · CONTINUOUS DETECTION VALIDATION

Detection Drift: Scheduled Runs Catch It the Same Week

Log sources update, agents get uninstalled, playbooks change. Weekly runs replay the same approved techniques and show when a detection that worked last week stops working.

TECHNIQUES DETECTED PER WEEKLY RUN · OF 24ILLUSTRATIVE
W1W2W3W4W5W6W7W8
DETECTION DRIFT2026 · week 42

Six detections went silent after a parser update.

A cloud log source update changed the event schema. The rules kept running and stayed green in the SIEM, but no longer matched.
DRIFT6 rules · user.arn null
SINCEparser update · 2026-10-14
07 · DETECTION RULE VALIDATION IN CI

Detection-as-Code: Validate Every Rule Change

When a detection changes in a pull request, Hayrok replays the techniques it is mapped to, including known variants, and posts expected vs observed back as a check.

detections/cloud/unusual_role_assumption.ymlcommit a1f3
title: unusual_role_assumption
match:
- event: AssumeRole
+ event: AssumeRole*
field: user.arn
severity: medium
attack: T1078.004
! Check failed · 1 of 3 variants firedhayrok / detection-validation
✓Rule changedpull request
✓Techniques mappedT1078.004
✓Variants replayed3 variants
✓Expected vs observedper variant
!Check postedfailed
T1078.004 VARIANTS REPLAYED
AssumeRole · console sessionalert fired
AssumeRole · CLI credentialsno alert
AssumeRoleWithWebIdentityno alert
08 · SAFE ATTACK EMULATION

Safe Detection Testing: Real Techniques in Production

Detection validation only means something on real systems. Every action is planned, approved, benign, and tagged so responders can separate test activity from an incident.

artefacts: benign

Benign emulation

Techniques run with harmless commands, canary secrets, and synthetic data. Nothing destructive runs.

runs: tagged

Tagged for your SOC

Every action carries a run ID, so analysts can tell test activity from a real incident.

scope: approved

Approved scope

Hosts, accounts, and techniques are fixed in a plan someone signs off before the run starts.

stop: anytime

Stop conditions

Anyone with authority can stop a run. Unexpected responses pause it automatically.

09 · WHO USES DETECTION VALIDATION

Who Uses Detection Validation: One Run, Four Teams

The same evidence record answers the engineer asking why a rule stayed silent, and the auditor asking whether monitoring controls operate.

DETECTION ENGINEERING

Know which rules work before an attacker tests them.

✓Find rules that are enabled but never fire✓Validate rule changes on every pull request✓Catch parser drift the week it happens
WHAT THEY READrule ID · failing stage · field See where detections fail →
10 · DETECTION VALIDATION VS SIEM TESTING

Detection Validation Compared: Check the Whole Path

SIEM rule tests replay sample logs into a rule. Breach and attack simulation shows whether a scenario was prevented. Detection validation runs the technique for real and checks every stage until someone sees the alert.

SIEM rule testing
Breach and attack simulation
Hayrok detection validation
WHAT IT TESTS
Rule logic against sample or replayed logs.
Whether simulated scenarios are prevented or detected.
Whether real techniques in your environment produce an alert someone sees.
TELEMETRY PATH
Skipped. Logs go straight into the rule.
Depends on agents and integrations.
Collection, parsing, rule, alert, and routing each checked.
WHEN IT FAILS
Pass or fail for the rule.
Prevented, detected, or not.
The first stage where the signal stopped, with the reason.
EVIDENCE
Test output.
Scenario report.
Expected vs observed for every step, in a replayable record.
AFTER A FIX
Rerun the unit test.
Rerun the scenario.
Replay the technique from the evidence record; the result joins the same record.
11 · FAQ

Detection Validation FAQ: What Evaluators Ask

It proves your detections work by running real attack techniques in your environment and checking what each layer blocked, logged, alerted, and routed. Each result is expected vs observed, with evidence.

Book a Detection Demo: See Which Detections Fire

Walk through connecting your SIEM, approving a technique set, running it, and reading expected vs observed for every stage.

✓Connect your SIEM and choose in-scope systems✓Approve a technique set and run it✓Read expected vs observed and retest after a fix
Book a Demo