Case Study

Using Message Routing Metadata to Improve Spam Detection

Case context: A support team is building a spam filter for a company inbox. The current detector only examines the words in the message body, but spammers keep rewriting their text so the filter misses many unwanted emails.

Question: Suggest one non-text feature idea that uses email metadata. What information should the team collect, and where should they look for it?

Sample answer: The team should use the email header, or envelope information, as the feature source. In particular, they should extract the list of mail servers that handled the message as it traveled to the inbox. That routing history gives a useful signal that does not depend on the email’s wording.

Key points:

  • Use metadata from the email header or envelope.
  • Extract the sequence or set of servers that processed the message.
  • Treat this as a non-content feature for spam detection.

Rubric: A correct response must name the email header or envelope as the source and state that the needed information is the set or sequence of servers the message passed through.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Machine Learning Strategy

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Yearning @ DeepLearning.AI