DeepFilter Email Security
DeepFilter Email Security is an AI-assisted secure email gateway (SEG) that runs entirely on your own servers. It sits in front of your existing mail infrastructure — Microsoft 365, Exchange, Google Workspace, Zimbra or Postfix — and scans inbound and outbound mail before delivery, quarantining phishing, business email compromise, ransomware and malicious attachments.
Its defining property is data sovereignty: routine email traffic and message content never leave your organization's boundary, and there are no third-party cloud API calls. The AI model runs on your server, not in a vendor cloud. This makes DeepFilter a fit for regulated sectors, public institutions and any organization whose compliance program (GDPR, KVKK and similar) makes third-party cloud inspection of email content difficult.
How it works: every message passes four stages — pre-analysis, deep inspection, evaluation and decision — across seven detection layers: SPF/DKIM authentication, blocklists (RBL/URIBL), statistical classification, an on-premise AI model, similarity analysis for near-duplicate and image-based campaigns, antivirus, and phishing/impersonation detection. The layers combine into a single trust score, and for every rule that fires the console shows what it is, why it triggered and how much it contributed — 400+ documented rules with explainable, per-rule scoring.
Key capabilities:
• Fully on-premise deployment on Rocky Linux 9 via signed RPM packages
• Regional threat-intelligence packs (the Türkiye pack is live: national cyber-security authority feed, signed and integrity-verified)
• Collective intelligence — a signature learned on one deployment propagates encrypted across the network without sharing email data
• Optional DLP module detecting region-specific data patterns (IBAN, national ID, card numbers) in outbound mail
• Quarantine management, allow/block lists, reporting and an end-user portal
• Management console and end-user portal in English and Turkish
Measured performance (vendor's own measurement, not a third-party benchmark): the AI model is trained on 200,000+ emails and reaches 99.44% precision and 94.65% recall, with a false-positive rate below 0.1%.
Licensing is based on scanned messages per day — not per mailbox or per seat — starting from 50 EUR + VAT per month for 1,000 messages/day. Public sector, education and non-profit organizations have separate pricing. Evaluation is a free proof of concept, installed and run end-to-end by the DeepFilter team on your own infrastructure.
DeepFilter is developed by FZM Bilgi Teknolojileri A.Ş., which holds an ISO/IEC 27001:2022 information security management system certificate (no. IFC-I-08-26-2987, accredited by TÜRKAK under the IAF MLA). The certified scope covers the development, delivery and support of email security software. The certificate applies to the management system, not to the product.
Note: unrelated to DeepFilterNet, an open-source speech noise-suppression project.
