Article

    Sample Datasets | Nightfall Documentation

    3 min read
    Last updated 1 month ago

    The following datasets can be used to test Nightfall's advanced AI-based detection capabilities. The data has been fully de-identified and can be used to test any data loss prevention (DLP) platform.

    PII Samples

    This dataset showcases Nightfall’s ability to detect Personally Identifiable Information (PII) with exceptional precision and minimal noise across text, spreadsheets, and screenshots. Samples include names, U.S. social security numbers, driver's license numbers, and more. See Image ID Samples for image samples of driver licenses and other ID types.

    This .ZIP contains the positive samples shown above, along with additional examples and negative lookalike samples for testing.

    PCI / Banking Samples

    This sample dataset demonstrates Nightfall's ability to detect sensitive banking and payment information with high precision and low noise in text, spreadsheets, and screen grabs. Samples include positive and negative examples of credit card numbers, routing numbers, IBAN codes, and SWIFT codes.

    This .ZIP contains the positive samples shown above, along with additional examples and negative lookalike samples for testing.

    API Keys

    Nightfall AI's fine-tuned API key detection LLM detects secrets with high precision and dramatically reduces false positives.

    This .ZIP contains the positive samples shown above, along with additional examples and negative lookalike samples for testing.

    Testing note: If a key status is marked as ‘Active’, please rotate the key immediately. Not all vendors provide an "Inactive" response code. In these cases or if the vendor service is offline, the finding status will be marked ‘Unverified’.

    Password Samples

    Nightfall AI detects passwords shared in conversational text and code.

    This .ZIP contains the positive samples shown above, along with additional examples and negative lookalike samples for testing.

    PHI Samples

    Nightfall’s PHI model surpasses traditional entity-based detectors by combining multiple signals — including PII and medical indicators — and analyzing their relationships and context to ensure only patient health–related content is flagged.

    This .ZIP contains the positive PHI samples shown above, along with additional examples and negative lookalike samples for testing.

    Crypto Key Samples

    This sample dataset demonstrates Nightfall's ability to detect cryptographic keys.

    This .ZIP contains the positive samples shown above, along with additional examples and negative lookalike samples for testing.

    Image ID Samples

    Nightfall’s computer vision (CV) transformer model outperforms legacy Optical Character Recognition (OCR) text scanning to identify driver’s licenses, passports, credit cards, and US social security cards even though images may be degraded (rotated, glossy, low contrast, blurry, skewed, or cropped).

    This .ZIP contains the positive samples shown above, along with additional examples.

    File Classifier Examples

    The File Classifier goes beyond entity detection by analyzing a document’s purpose, structure, format, and contextual signals to accurately identify intellectual property, proprietary source code, and other sensitive internal records. These examples demonstrate how the classifier protects confidential materials—including internal source code, legal and regulatory drafts, HR documents, and strategic planning files—across a wide range of real-world scenarios.

    All Sample Datasets

    This ZIP file includes all positive and negative lookalike samples across PII, PCI, Banking, PHI, credentials, and image-based datasets. It’s designed to help you evaluate Nightfall’s detection precision and compare performance in your DLP proof of value (POV) testing.