Phishing Email Clustering Using Quasi-Ground-Truth Machine Learning

Datasets referenced in a conference paper (submitted)

Zip files are password-protected; the password will be revealed after paper acceptance.

phish_public - phishing messages, a dataframe with columns:

• ts_anon - timestamp
• from_anon - sender
• to_anon - recipient
• subject_anon -suject
• content_anon - message text
• client_ip_anon - IP of the sender
• matched_domains - matched known malicious domains
• cluster_gt - spam campaign ID (ground truth)
• emb - embeddings of URLs in message

phish_public_teaser is the preview of the file 

test_public - test dataset, a dataframe with columns:

• msg_id - message ID, as in the original dataset: AISec_Clustering
• paper_cluster_id - ground-truth
• emb - embeddings of URLs in message
• siamese_embeddings - embeddings of message text