Case context: You are building a message-filtering service for a small business platform and need many examples of scam messages. A list of scam sender groups is already known, but your team does not yet have enough labeled examples to train a classifier.

Question: How can you use a honeypot approach to automatically gather a training set from these known scam sources?

Sample answer: Create a set of unused decoy inboxes and place them where the scam sender groups are likely to reach them, such as in public sign-up forms or leaked contact lists. Because no real users monitor these inboxes, any messages they receive can be collected automatically and treated as scam examples, producing a large labeled dataset with little manual review.

Key points:

• Create decoy inboxes.
• Expose the decoys to known scam sources.
• Automatically collect the incoming messages.
• Label the collected messages as scam training data.

Rubric: The response must explain creating decoy email inboxes, exposing them to known scam sources, and automatically collecting and labeling the incoming messages as scam examples.