Changelog

Each DLP entity type page now includes example matches, false positive guidance, and recommended enforcement actions. Documentation is generated from the validation dataset, so examples reflect actual detection behavior — not hypothetical patterns.

New section in the trust center explaining how Arbitex measures and reports DLP detection accuracy. Covers the validation dataset methodology, per-entity accuracy metrics, and how to request accuracy data for specific entity types relevant to your compliance requirements.

The DLP feature page now describes how detection quality is built for each detector category — checksum validation for structured formats, contextual signals for context-dependent patterns.

New engineering blog post explaining how we build the comprehensive validation dataset, measure per-entity-type detection accuracy, run isolated versus integrated testing, and gate CI on regression. Published so security teams evaluating DLP products can ask their vendors the same questions.

Fixed a timing-dependent test failure in the SSO session handoff flow. The test relied on a session expiry window that could race under CI load. Replaced the fixed timeout with a deterministic session state check, eliminating intermittent failures in the identity test suite.

Optimized the admin portal list views for large deployments. Detection log, user activity, and audit trail pages now use virtualized rendering, reducing initial load time and memory usage when scrolling through high-volume event histories.

Outpost data plane deployments now receive the same DLP pattern updates as the cloud-hosted pipeline within the same release window. Pattern versioning ensures air-gapped Outpost instances can verify they are running the current detection ruleset.

Each of the 65+ DLP entity types now has a dedicated documentation entry describing what it detects, which checksum or format validation it applies, which compliance frameworks reference it, and what enforcement actions are available. Accessible from the admin portal and the developer documentation hub.

New accuracy validation view in the admin portal showing per-detector accuracy metrics across production traffic. Security teams can identify which detectors are producing false positives and tune thresholds without guessing — measured accuracy, not assumed accuracy.

The DLP pipeline now classifies detected secrets into 39 distinct entity types — API keys, OAuth tokens, database credentials, cloud access keys, private keys, and more. Each type carries its own severity, compliance mapping, and enforcement action, replacing the previous single "secret" category with granular classification.

The admin portal now enforces role-based access control with four built-in roles: Viewer, Analyst, Admin, and Super Admin. Each role defines which portal sections, configuration actions, and audit records a user can access — least-privilege by default, with custom role support for enterprise deployments.

Strengthened Content Security Policy headers across the admin portal, added rate limiting to authentication endpoints, and enforced secure cookie attributes on all session tokens. These changes reduce the attack surface for XSS, credential stuffing, and session hijacking.

New /security/trust page providing a single reference for the platform's security posture — encryption standards, audit logging architecture, compliance framework coverage, identity stack, and deployment security controls. Designed for security teams conducting vendor assessments.

Fixed three regex patterns that were over-matching in conversational AI content: date-like strings no longer trigger SSN detectors, code variable names no longer trigger API key detection, and markdown formatting no longer triggers credit card patterns. Combined false positive rate reduced measurably across production traffic.

Outpost deployments now support graceful drain mode for zero-downtime upgrades. When a drain signal is received, the data plane stops accepting new requests, completes in-flight inspections, and shuts down cleanly — preserving audit chain integrity throughout.