Nuevo v1.2.83

Plexicus v1.2.83: Hardened AI Pentest Isolation, Smarter Retest Scope & SSO That Just Works

AI Pentest engine pods are fully isolated from platform storage and cluster credentials, retest runs no longer re-attack excluded domains, and SSO login, configuration, and role management get a comprehensive hardening pass.

🚀 What’s New

  • Retest Respects Excluded Domains: When you retest a previously scanned target, AI Pentest now honors the domain exclusions you set in the scope wizard — a retest will never re-attack a host you explicitly removed from scope.
  • Cumulative Status Reports for Linked Pentests: A linked pair of AI Pentest runs now produces both individual closed reports and a single cumulative status report across the pair, giving you a continuous view of an application’s security posture over time.
  • Dismissed-Threat Recheck on Changed Code: Previously dismissed threats are now automatically re-checked only where their underlying code has changed, so you can trust that a dismissal stays valid until the threat surface actually shifts.
  • Prior-Map Revalidation: When you launch a continue or linked pentest, Plexicus now revalidates the stored application map against the current target definition before the run begins, so you’re alerted if the application’s identity has drifted since the last engagement.

🛠 Improvements

  • Findings Deep-Link Reliability: Opening a finding from a shared or bookmarked link now seeds the surrounding list filter correctly — prev/next navigation stays enabled and your filter context is preserved as you move between findings.
  • Finding Remediation Panel Accuracy: The remediation panel on a finding detail now tracks the finding you’re actually viewing as you navigate prev/next in place, so remediation actions and status always match the correct finding.
  • SSO Configuration Without Re-Entering Secrets: The SSO Advanced page now saves and dry-runs your OIDC configuration using the stored client secret — you no longer need to re-enter the secret on every save or test.
  • SSO Link Page Redesign: The first-login SSO account-linking page now matches the main login layout, branding, and styling for a seamless authentication experience.
  • Partner Programme API: Partners can now view their programme cycle, tier, statements, and deals, and manage certification seats through a dedicated partner-facing API surface.
  • Repository Scan Endpoint in API Reference: The trigger-scan endpoint is now visible in the generated OpenAPI reference, making it easier to discover and integrate.
  • Retest Client Routes: Clients can now read their retest run history and details through dedicated API routes, with failure reasons presented in clear, client-safe language.

🔒 Security

This release includes a comprehensive hardening pass across AI Pentest isolation, authentication, and access control:

  • AI Pentest engine pods no longer hold any object-storage credentials or cluster API tokens — all artifact read/write now flows through an engagement-scoped broker with signed requests, so AI agents can never access other tenants’ data or platform infrastructure.
  • AI SAST scanning pods no longer receive object-storage root keys when running tenant code analysis.
  • GitHub OAuth login state is now bound to the originating browser, preventing login CSRF attacks, and session tokens are no longer exposed in redirect URLs.
  • Auth cookies now carry SameSite=Lax attributes from a single source, and session tokens are no longer sent to third-party endpoints.
  • Role assignment is now ceiling-enforced: a caller can never bind, invite, or map a role above their own level, and custom roles can no longer self-escalate.
  • SSO and SCIM provisioning now write the access bindings that enforcement reads, so a just-in-time SSO user or an IdP group change takes effect immediately.
  • Azure DevOps connector rejects stored organization URLs pointing at internal hosts, preventing SSRF via manipulated connector configuration.
  • Platform API keys stored in BYOAI slots or registry rows are never served to tenant-configured hosts, closing a residual key-exposure path.
  • Dependency and framework packages have been updated to address known vulnerabilities.

🔧 Fixes

  • Pentest Cancellation Cleanup: Cancelling an AI Pentest run now fully cleans up its stored artifacts — no leftover screenshots or report objects remain after cancellation.
  • Pentest Reservation Release: A pentest reservation is now released only after the engine is confirmed stopped, preventing orphaned runs from holding quota.
  • Duplicate Finding Deduplication: AI Pentest results that describe the same vulnerability with different phrasing are now collapsed into a single finding, so your findings list stays clean.
  • Pentest Scope Host Ordering: A web target’s own host now always appears first in the stored allowed-domains list, preventing scope misconfiguration that could exhaust the run budget.
  • Blank Scope Entries Stripped: Empty or whitespace-only entries in allowed-domains are now silently dropped, preventing scope parsing errors.
  • SSO Browser Login Fixed: Browser-based SSO login no longer fails silently — the session token is now correctly sent on the immediate post-login verification call.
  • Self-Serve Subscription Renewals Filter: The renewals view now correctly includes self-serve subscriptions when sorting and filtering by end date.
  • Retest Cancellation Refund: Cancelling a retest run now deterministically heals the pending refund reservation instead of leaving it held.
Calificación

Comprueba si el AI Swarm Pentest encaja en tu entorno.

Déjanos el contexto mínimo. Revisaremos el alcance y te indicaremos el siguiente paso comercial.

Antes de enviar — verifica que encajas

0 / 280

Sin compromiso. Si no encajas, te lo decimos.

SAMPLE HANDOVER · ILLUSTRATIVE

Sample evidence handover

A trimmed view of what your team receives at the end of an AI Swarm Pentest engagement. Real engagements include full technical evidence, executive narrative, and a remediation plan.

VALIDATED FINDING Evidence attached

Server-Side Request Forgery in webhooks/receiver

demo-project/sample-app · src/webhooks/receiver.py:42

SeverityHigh CVSS 3.18.6 Priority79 Confirmedvia replay

Untrusted caller-supplied URLs reach an internal egress without an allowlist. Replayed in a sandbox against a fresh authorized target — the same control was validated to fail twice.

REVIEWER-READY REMEDIATION Merge-ready PR

Validate the target URL against an allowlist of permitted hostnames. Reject private/internal IP ranges. Enforce HTTPS only.

plexicus/remediation/webhooks-ssrf 3 changed · 0 new files
42resp = requests.get(target_url)
42+if not is_allowed_host(target_url):
43+  raise WebhookRejected(target_url)
44+resp = requests.get(target_url, timeout=5)
Every engagement hands over:
  • Executive briefing
  • Validated findings list
  • Merge-ready PRs
  • Compliance mapping (NIS2 · DORA · CRA)
Ronda privada Para inversores