CPM-CC Email Metadata Dataset: 109 Emails for Cross-Platform Phishing Detection

This dataset supports the CPM-CC (Cross-Platform Metadata Consistency Check) algorithm described in the paper "Data-Driven Phishing Email Detection by Analyzing Metadata Across Platforms for Enhanced Security" (SSRN ID: 5189807; Zenodo DOI: 10.5281/zenodo.19564559). The dataset contains metadata features extracted from 109 emails — 62 phishing and 47 legitimate — across three categories: 23 real-world emails from a personal inbox, 10 high-evasion phishing samples that passed SPF/DKIM/DMARC authentication but were verified as phishing through cross-platform metadata analysis, and 76 pattern-validated synthetic emails based on documented attack signatures. Features include SPF/DKIM/DMARC authentication results, relay path analysis, Unicode/Punycode domain spoofing detection, suspicious via header flags, return-path mismatch indicators, and composite CPM-CC risk scores. Four ML classifiers (Decision Tree, Random Forest, Neural Network, SVM) were trained and evaluated on this dataset, achieving up to 98.9% accuracy. See the included README for full schema and methodology documentation.