PhishRadar

Detect phishing/scam domains from Certificate Transparency logs. Can be useful for internal security teams for monitoring company-specific domains, as well as for sectorial/national CERTs for monitoring the domains of their constituents.

Logs are retrieved via Certstream. Keywords are set in a YAML configuration file and detected via two methods:

• for common subwords, results from WordSegment are scanned
• for rarer words (e.g. brand names), we simply check if the domain contains the keyword - this is because WordSegment may not correctly separate unknown subwords.

For better customization, a minimal threshold of matching keywords can be set.

The idea was born after encountering phishing attacks and malware Command-and-Control communication involving domains impersonating Armenian government bodies (1, 2).

Installation and usage

pip install phishradar
phishradar --config ./config.yaml

Sample configuration

certstream_url: wss://certstream.calidog.io/
keywords:
  - test
  - keyword
  - yourbrand
whitelist:
  - exclude
  - these
threshold: 1  # Number of matched keywords to alert for
output:
  console: True
  file: output.log
  # Example use case of a webhook sink - a Telegram bot for alerting
  # The following templates are supported in messages (URL and body):
  #   $domain - defanged domain name
  #   $matches_list - bullet point list of matched keywords
  #   $matches_comma - comma-separated string of matched keywords
  webhook:
    url: "https://api.telegram.org/bot<INSERT_YOUR_BOT_TOKEN>/sendMessage"
    body: '{"chat_id": "<INSERT_YOUR_CHAT_ID>", "parse_mode": "html", "text": "<b>$domain</b> contains: <i>$matches_list</i>"}'

Further work

• Try out alternative threshold mechanism (e.g. weighted keywords)
• Experiment with word segmentation via a Small Language Model