{
    "schema_version": 1,
    "generated_at": "2026-07-26T15:57:23.483964+00:00",
    "product": {
        "name": "Microsoft Presidio",
        "version": "2.2.364",
        "configuration": "Default recognizers with en_core_web_sm; score threshold 0.0",
        "source": "https://github.com/microsoft/presidio"
    },
    "corpus": {
        "name": "RedactCloud Structured Redaction Corpus",
        "version": "1.0.0",
        "sha256": "0853d072502092a14aa533970b7e6a6d3039f892ade075744f9a586f960c7e37",
        "case_count": 46,
        "positive_entity_count": 25,
        "shared_types": [
            "CREDIT_CARD",
            "EMAIL",
            "IBAN",
            "IP_ADDRESS",
            "PHONE_US",
            "SSN"
        ],
        "scope": "Only entity types supported by both default Presidio and RedactCloud detectors; all clean cases are included."
    },
    "accuracy": {
        "true_positives": 23,
        "false_positives": 2,
        "false_negatives": 2,
        "precision": 0.92,
        "recall": 0.92,
        "f1": 0.92,
        "clean_case_count": 25,
        "clean_case_pass_rate": 0.96,
        "failed_case_count": 4,
        "by_type": {
            "CREDIT_CARD": {
                "true_positives": 6,
                "false_positives": 0,
                "false_negatives": 0,
                "precision": 1,
                "recall": 1,
                "f1": 1
            },
            "EMAIL": {
                "true_positives": 5,
                "false_positives": 0,
                "false_negatives": 0,
                "precision": 1,
                "recall": 1,
                "f1": 1
            },
            "IBAN": {
                "true_positives": 4,
                "false_positives": 0,
                "false_negatives": 0,
                "precision": 1,
                "recall": 1,
                "f1": 1
            },
            "IP_ADDRESS": {
                "true_positives": 3,
                "false_positives": 0,
                "false_negatives": 0,
                "precision": 1,
                "recall": 1,
                "f1": 1
            },
            "PHONE_US": {
                "true_positives": 3,
                "false_positives": 1,
                "false_negatives": 0,
                "precision": 0.75,
                "recall": 1,
                "f1": 0.8571
            },
            "SSN": {
                "true_positives": 2,
                "false_positives": 1,
                "false_negatives": 2,
                "precision": 0.6667,
                "recall": 0.5,
                "f1": 0.5714
            }
        },
        "failed_case_ids": [
            "ssn-dashed",
            "ipv4-private",
            "mixed-customer",
            "clean-ssn-area-nine"
        ]
    },
    "limitations": [
        "This is a RedactCloud-run comparison, not an independent audit.",
        "The shared subset excludes secret types that default Presidio does not advertise as built-in entities.",
        "The corpus is synthetic and maintained by RedactCloud.",
        "Presidio results depend on recognizer configuration, NLP model, language, and score threshold."
    ]
}