phish-messages / data /evaluation.json
alirezaaminzadeh's picture
Publish Phish synthetic messages and campaign split (part 2)
007bcd4 verified
Raw History Blame Contribute Delete
1.51 kB
{
"n_test": 96,
"split": "campaign-held-out",
"never_outputs_safe": true,
"evidence_link_rate": 1.0,
"hard_negative_suspicious_rate": 0.0,
"decision_counts": {
"suspicious": 48,
"needs_review": 16,
"insufficient_evidence": 32
},
"methods": {
"rule-structural": {
"pr_auc": 1.0,
"brier": 0.10375000000000001,
"recall_at_fpr_0_05": 1.0,
"recall_at_fpr_0_10": 1.0,
"mean_score": 0.34166666666666673,
"pr_curve_points": 6
},
"tfidf-lr": {
"pr_auc": 1.0,
"brier": 0.14890835989583331,
"recall_at_fpr_0_05": 1.0,
"recall_at_fpr_0_10": 1.0,
"mean_score": 0.629509375,
"pr_curve_points": 43
},
"lightgbm-structural": {
"pr_auc": 1.0,
"brier": 0.16022288833333334,
"recall_at_fpr_0_05": 1.0,
"recall_at_fpr_0_10": 1.0,
"mean_score": 0.7102166666666667,
"pr_curve_points": 6
}
},
"by_language": {
"en": {
"n": 64,
"pr_auc_lightgbm": 1.0,
"recall_at_fpr_0_10": 1.0
},
"fa": {
"n": 32,
"pr_auc_lightgbm": 1.0,
"recall_at_fpr_0_10": 1.0
}
},
"limitations": [
"Synthetic fixtures only. Lab PR-AUC is not operational phishing accuracy.",
"Copied Authentication-Results headers are not treated as proof.",
"URLs are parsed and defanged; they are never fetched.",
"A similar logo or a new domain is not enough for a malice verdict."
]
}