Malicious Domain Protection

Training the machine learning model

To train a machine learning classifier, training data is needed. DNSFilter's data scientists had access to the world's fastest Dual-Anycast resolvers that process over 89 billion DNS queries daily for over 30 million users.

Over 35 million daily DNS queries are identified as threats and blocked, before they ever resolve to a client machine. Malicious Domain Protection's classifier was trained and fine-tuned using over 16 million queries, a mix of both known good and known bad domains contributing to a highly accurate machine learning model.

The threats we face

Unknown malicious domains seen in the wild can be used with malware, botnets, phishing attacks or command-and-control servers.

Attackers use the period before their domain is discovered to launch attacks that go undetected by firewalls, IDS, and other threat detections that rely on threat feeds or other information.

Some malware families algorithmically generate upwards of 250,000 unique domains. Also known as DGAs, these domains are often the first signs of malware, spambots, phishing, and cryptojacking, and more.