Skip to content

SECURITY_AUDIT.md

Latest commit

 

History

History
84 lines (61 loc) · 4.34 KB

SECURITY_AUDIT.md

File metadata and controls

84 lines (61 loc) · 4.34 KB
Raw

BLIP Security Audit Report

Date: 2026-03-10 Scope: Full-stack (Web: Next.js, Mobile: Flutter, DB: Supabase)

Executive Summary

BLIP is an end-to-end encrypted ephemeral communication platform. Overall security posture is solid, with proper E2EE implementation (Curve25519 + XSalsa20-Poly1305), parameterized database queries, and comprehensive rate limiting. This audit identified 4 code-level vulnerabilities (now fixed) and several configuration recommendations for production hardening.

Overall Score: 7.5/10 (8.5/10 after fixes applied)

Fixed Vulnerabilities

1. [CRITICAL] Missing CSRF Protection — board/batch-status

  • File: web/src/app/api/board/batch-status/route.ts
  • Issue: checkOrigin() was not called, allowing cross-origin CSRF attacks
  • Fix: Added checkOrigin() validation and parseJsonBody() safe parsing

2. [CRITICAL] Missing Array Element Type Validation — board/batch-status

  • File: web/src/app/api/board/batch-status/route.ts
  • Issue: Array.isArray(boardIds) only checks container type, not element types. Non-string elements could reach database queries
  • Fix: Replaced with isStringArray(boardIds) which validates each element

3. [HIGH] Weak Randomness Fallback — ChatRoom

  • File: web/src/components/chat/ChatRoom.tsx:99-101
  • Issue: Math.random() used as fallback for transfer ID generation (predictable, not cryptographically secure)
  • Fix: Replaced with crypto.getRandomValues() for cryptographically secure random bytes

4. [HIGH] Unsafe parseInt Without Range Validation — upload-image

  • File: web/src/app/api/board/upload-image/route.ts:135-137
  • Issue: parseInt() can produce NaN or arbitrarily large values for width/height/displayOrder
  • Fix: Added bounds clamping (width/height: 0-10000, displayOrder: 0-100) with NaN fallback to 0

5. [HIGH] Missing Origin Check — upload-image

  • File: web/src/app/api/board/upload-image/route.ts
  • Issue: POST endpoint missing checkOrigin() CSRF protection
  • Fix: Added checkOrigin() validation

Remaining Recommendations (Not Auto-Fixed)

HIGH Priority

# Issue Location Recommendation
1 CSP allows unsafe-inline/unsafe-eval web/next.config.ts:38-39 Remove unsafe directives; sandbox ad scripts in iframe
2 Sensitive data in localStorage web/src/hooks/useBoard.ts Move encryption keys/admin tokens to sessionStorage or memory-only
3 Ad script injection via dangerouslySetInnerHTML web/src/components/MonetagAds.tsx Add Subresource Integrity (SRI); verify ad network legitimacy

MEDIUM Priority

# Issue Location Recommendation
4 PBKDF2 iterations below NIST 2023 recommendation web/src/lib/crypto/keys.ts:49 Upgrade from 100,000 to 120,000+ iterations
5 No certificate pinning (mobile) mobile/lib/core/network/api_client.dart Implement cert pinning for Supabase and API domains
6 Missing iOS ATS configuration mobile/ios/Runner/Info.plist Add explicit NSAppTransportSecurity policy
7 Rate limiting disabled in development web/src/lib/rate-limit.ts:46-49 Ensure NODE_ENV is never 'development' in production

LOW Priority

# Issue Recommendation
8 No MIME type whitelist for uploads Validate file content type server-side
9 Console.error may expose DB details Sanitize error logs in production

Security Strengths

  • No hardcoded secrets — all credentials properly externalized via environment variables
  • Strong E2EE — Curve25519 ECDH + XSalsa20-Poly1305 (NaCl), consistent across web and mobile
  • SQL injection protected — all queries use parameterized Supabase client SDK
  • Comprehensive rate limiting — applied to all sensitive endpoints via Upstash Redis
  • Row Level Security — Supabase RLS properly configured
  • HTTP security headers — HSTS, X-Frame-Options, X-Content-Type-Options, Referrer-Policy
  • Screenshot/screen recording protection (mobile)
  • Proper CSRF protection — origin validation on 31/31 POST endpoints (after fixes)
  • Secure key derivation — PBKDF2 with salt for password-based keys
  • No dangerous code patterns — no eval(), exec(), or unsafe deserialization