Security at DocuPDF

Uploaded files are written to a temporary directory on our processing server. Once the operation is complete and the download is served, the file is automatically deleted. Our cleanup job runs every 60 seconds.

• Temporary files auto-deleted after processing
• No backups of your uploaded files
• No file metadata stored in our database

All traffic between your browser and DocuPDF is encrypted using HTTPS / TLS 1.3. We use Cloudflare as our edge provider, which enforces HTTPS and strips insecure connections.

• TLS 1.3 for all connections
• HSTS enforced at the edge
• No unencrypted HTTP accepted

DocuPDF runs behind Cloudflare, which provides DDoS mitigation, Web Application Firewall (WAF) rules, bot protection, and IP reputation filtering — before traffic ever reaches our servers.

• DDoS mitigation (layers 3, 4, and 7)
• WAF with OWASP ruleset enabled
• Bot traffic filtered at edge

Passwords are hashed using bcrypt (never stored in plain text). JWT tokens are short-lived and signed with a server-side secret. We enforce brute-force lockouts after repeated failed login attempts.

• bcrypt password hashing (cost factor 12)
• Short-lived JWT tokens
• Lockout after 5 failed logins in 15 minutes
• IP + email reputation checks on signup

Files are processed programmatically by the PDF engine (PyMuPDF, pikepdf). No human ever opens or reviews your documents. AI features process extracted text — raw files are never sent to an LLM.

• Processing is fully automated
• No human review of file contents
• AI tools receive text only, not raw files
• OCR handled by an isolated external microservice

Every uploaded file is validated against its magic bytes before processing. Files that claim to be PDFs but aren't are rejected immediately. This protects both you and our servers from malformed or malicious payloads.

• Magic byte validation on every upload
• Extension and MIME type checks
• File size limits enforced server-side

OCR text extraction runs on a dedicated, isolated external server — completely separate from the main application. Files sent for OCR are processed and discarded immediately; they never touch our primary database or file storage.

• OCR runs on a separate hardened VPS
• Authenticated API calls with secret key
• No OCR file retention — processed and gone