Classifiers

How Bigeye Detects Sensitive Data

Bigeye Sensitive Data Scanning uses a combination of pattern matching, checksum validation, and machine learning to detect sensitive data across your structured data sources. These techniques work together to build confidence before a finding is produced, balancing precision with recall to minimize false positives.

All detection runs entirely within your environment. Bigeye does not train models on your data — classification is performed using inference only.

Key Concepts

Classifier: A classifier contains the detection logic used to identify a specific type of sensitive data. Each classifier produces a single data class as its output when a match is found (e.g., "SSN", "Email Address", "Credit Card Number"). When configuring a classifier, you will select the data class to be applied when the classifier's detection patterns are satisfied.

Data class: A label that describes the type of sensitive data discovered. Each data class can be mapped to a sensitivity level — Public, Internal Only, Confidential, or Restricted — to indicate the level of risk associated with that data.

Detector: The atomic unit of detection logic within a classifier. Detectors can use regex patterns or machine learning models. When a classifier contains multiple detectors, they can be combined using AND or OR logic.

Out-of-the-Box Classifiers

Bigeye ships with a curated set of out-of-the-box classifiers ready to use in your scan jobs. These cover common sensitive data types across PII, PHI, PCI, and financial data.

Personally Identifiable Information (PII)

Data Class	Description
Name	Full or partial person names
Email Address	Email addresses
Phone Number	Phone numbers (USA only)
Home or Mailing Address	Physical street and mailing addresses
Date of Birth	Dates of birth
Place of Birth	Location of birth
Age	Age values
Mother's Maiden Name	Maiden name identifiers, commonly used in security verification
Nationality, Religion, or Political Group	Demographic and group affiliation identifiers
Password	Password fields identified by column name and value patterns
Personal Account Usernames	User account identifiers
IP Address	IPv4 and IPv6 addresses
Location	General geographic or location data
Date or Datetime	Date and timestamp values
Web URL	Web addresses that may contain identifying information
Device Identifier or Serial Number	Hardware or device serial numbers
Crypto Wallet ID	Cryptocurrency wallet addresses

Government & Travel Identifiers

Data Class	Description
US SSN/TIN	US Social Security Numbers and Tax Identification Numbers
US ITIN	US Individual Taxpayer Identification Numbers
US Driver's License Number / State ID	US driver's license and state ID numbers across all US states
US Passport Number	US passport numbers
Vehicle Identifiers (VIN, Plate #, Registration #)	Vehicle identification numbers, license plates, and registration numbers

Protected Health Information (PHI)

Data Class	Description
Patient Name	Patient name fields in healthcare contexts
MRN (Medical Record Number)	Medical record number identifiers
Diagnosis	Diagnostic codes and descriptions (e.g., ICD codes)
Treatment Codes	Medical treatment and procedure codes
Medical Information	General medical and clinical information
Medical License	Medical practitioner license numbers
Provider NPI (National Provider Identifier)	National Provider Identifier numbers
Health Plan / Insurance Number	Health insurance and plan identifiers
Healthcare Admission Date	Patient admission dates
Healthcare Discharge Date	Patient discharge dates

Payment Card Industry (PCI)

Data Class	Description
Credit Card Number	Credit and debit card numbers, validated with checksum
CC Expiration Date	Card expiration dates
CVV	Card verification values

Personal Financial Information (PFI)

Data Class	Description
US Bank Account Number	US bank account numbers
IBAN Code	International Bank Account Numbers
Account PIN	Account PIN codes
Credit Score	Credit score values

If the above list lacks a classifier that you need, please inform your Bigeye representative.

Customizing Classifiers

Clone and Adjust

You can clone any out-of-the-box classifier to create your own version. Cloned classifiers can be adjusted — for example, modifying or adding regex patterns to better match data formats specific to your organization.

Build from Scratch

You can also create classifiers entirely from scratch. When building a custom classifier, you can add one or more detectors using regex patterns, machine learning models, or a combination of both. Multiple detectors within a single classifier can be linked with AND or OR logic to fine-tune detection accuracy.

When creating or editing any classifier, you will assign a data class that will be applied as the finding output whenever the classifier's detection logic is satisfied.