SOC 2 endpoint logging and monitoring controls
Last Updated: September 2, 2026
ET Ducky is not SOC 2 audited. This page states that first because it is the question most reviewers arrive with, then documents how the platform's controls map to the Trust Services Criteria and what evidence it produces for an auditor working on your report.
Attestation status
ET Ducky holds no SOC 2 report, Type I or Type II. Readiness work is in progress. The security whitepaper documents how controls map to the SOC 2 Trust Services Criteria, ISO 27001 Annex A, NIST CSF and GDPR Article 32. A mapping is a self-assessment performed by the vendor. It is not an attestation, it has not been examined by an independent auditor, and it is not offered as a substitute for one.
Every third-party service the platform depends on does hold a current attestation. That is a real fact about the supply chain and a limited one about ET Ducky itself.
| Sub-processor | Purpose | Attestation |
|---|---|---|
| Clerk | Authentication and identity | SOC 2 Type II |
| Stripe | Payment processing | PCI DSS Level 1 |
| Anthropic | AI analysis engine | SOC 2 Type II |
| SendGrid | Transactional email | SOC 2 Type II |
| DigitalOcean | Infrastructure hosting | SOC 2 Type II, ISO 27001 |
| Cloudflare | DNS, CDN, DDoS protection | SOC 2 Type II, ISO 27001 |
Control mapping
This is the mapping as it appears in the whitepaper, reproduced here so a reviewer does not have to open a 40-page document to find it.
| Control area | Implementation | Criteria |
|---|---|---|
| Access control | Per-agent bearer tokens, Clerk role-based access control, organization isolation, admin endpoints restricted to localhost | CC6.1, ISO 27001 A.9, NIST AC |
| Encryption in transit | TLS 1.2 or higher on all connections, WSS for WebSocket | CC6.7, ISO 27001 A.10, NIST SC-8 |
| Encryption at rest | AES-256 database encryption, AES-256-GCM token encryption | CC6.7, ISO 27001 A.10, NIST SC-28 |
| Audit logging | Structured logging of authentication events, agent connections, alert evaluations, command risk assessments and execution decisions | CC7.2, ISO 27001 A.12.4, NIST AU |
| Data minimization | System telemetry plus the agent's connection IP; no end-user content, documents or keystrokes; configurable retention with automatic purge | P6.1, GDPR Art. 5(1)(c) |
| Secrets management | Environment variables with restricted file permissions, no hardcoded secrets | CC6.1, NIST IA-5 |
| Vulnerability management | Signed, hash-verified agent auto-update, container patching, managed database patching | CC7.1, ISO 27001 A.12.6, NIST SI-2 |
| Threat detection and response | Always-on behavioral monitor, automatic network isolation on high-confidence detection, HMAC-TOTP local lift codes, automatic ticket creation with evidence | CC7.2, CC7.3, ISO 27001 A.12.4, A.16, NIST IR-4, SI-3 |
| Incident response | Targeted containment with per-incident evidence report and review ticket, token revocation, credential rotation, command risk scoring | CC7.3, ISO 27001 A.16, NIST IR |
| Third-party risk | All third-party services hold current attestations, data sharing minimized | CC9.2, ISO 27001 A.15, NIST SA-9 |
| Network security | AllowedHosts restriction, CORS policy, VPC database isolation, no inbound agent ports, geographic IP restrictions on registration and downloads | CC6.6, ISO 27001 A.13, NIST SC-7 |
CC7.2: what monitoring evidence looks like
CC7.2 asks that the entity monitor system components for anomalies indicative of malicious acts, natural disasters and errors, and that it analyze them to determine whether they represent security events. Behavioral detection against kernel telemetry is a direct fit, because the evaluation runs on events the operating system emitted rather than on a report from the thing being monitored.
Ten always-on rules run on both Windows and Linux: mass file access, mass file rename to ransom extensions, ransom-note patterns, shadow copy and backup deletion, suspicious execution chains, reverse-shell heuristics, privilege escalation from non-interactive parents, unusual outbound traffic from system daemons, a crypto-miner heuristic, and a kill-chain rule that fires when two or more of the others occur within five minutes.
Two further mechanisms produce anomaly evidence and both ship default-off with a shadow mode, which is itself useful to an auditor because it shows tuning before enforcement:
- Approved-app binary integrity. Agents report a SHA-256 of each approved application's primary binary. The cloud holds a per-organization ledger with trust on first use, the reference elected by agent majority, and any later divergent hash for the same application, version and architecture raises a detection once.
- Fleet behavioral baselines. The cloud learns a per-application profile from the organization's own fleet in 15-minute rollups and flags outlier hosts.
CC6.1: logical access and tenant isolation
Isolation runs in two layers on every database read and write. Application-layer query filters add the organization predicate to every query against tenant-scoped entities, so a controller cannot omit it. Database-layer row-level security policies enforce the same predicate in PostgreSQL, so a query that escapes the application filter is still rejected.
Lookups that must run before tenant context exists, such as bearer token validation and webhook handlers, opt in to a documented bypass scope and are flagged by an in-tree static analyzer if introduced without justification. There is no administrative path to cross-tenant access. Admin endpoints are bound to localhost and provide no cross-organization view.
At the console layer, identity is handled by Clerk. ET Ducky does not store passwords or manage sessions. TOTP, SMS and passkey factors are available, and SAML or OIDC single sign-on is available on Clerk's enterprise plan.
CC7.3 and CC7.4: evaluating and responding to events
On a detection at or above the confidence threshold, containment is automatic and scoped to the implicated endpoint. A ticket is opened in the connected ticketing system carrying the alert evidence and a note that isolation was applied, which gives the incident a dated record without an operator present.
Release is separately attributable. Lifting isolation requires an authenticated dashboard request plus independent approval from a person at the device, through either a single-use HMAC-TOTP code or an on-device approval prompt with a five-minute window. A stolen dashboard session cannot lift containment on its own.
Data belonging to an organization under a security hold is exempt from the retention purge, so evidence survives an investigation that runs longer than the retention window.
Using ET Ducky as evidence in your own audit
A vendor without its own report can still be the source of evidence for your report. What your auditor tests is the completeness and integrity of records in your environment. The records most often useful:
| Criterion | Record available |
|---|---|
| CC6.1 | Authentication events, agent connection events, organization-scoped access |
| CC6.8 | Approved-app binary hash ledger and divergence detections |
| CC7.1 | Agent update records with signature and hash verification, patch rollout and post-patch verification results |
| CC7.2 | Alert rule evaluations per agent with timestamps, behavioral detections with triggering evidence |
| CC7.3 | Isolation actions, the ticket raised, and the approval path used to lift |
| CC8.1 | Script creation and modification records with SHA-256 content hash, command executions with risk score and decision |
Where the vendor's own status does matter is CC9.2, vendor and business partner risk. An auditor will ask what due diligence you performed on a vendor that runs privileged code on your endpoints and holds no report of its own. The whitepaper, the security posture page and a completed questionnaire response are what that file is built from.
What the mapping is not
It is not an attestation, it has not been examined by an independent auditor, and it covers only the criteria a platform can address. Nothing here speaks to CC1 through CC5, which cover the control environment, communication, risk assessment, monitoring of the control environment, and control activities at the organizational level. Those are governance and no product supplies them.
Related: cyber insurance endpoint control questions and GLBA and FFIEC endpoint monitoring evidence.
Reviewing ET Ducky as a vendor?
The whitepaper answers most of a standard questionnaire already. If yours asks something it does not cover, send it and get a written answer, including the items where the answer is that the control does not exist yet.
Book a 25-minute review call Request the questionnaire response
Questions about a specific control go to [email protected]. The security whitepaper is the source of record for everything on this page.