DataStealth Data Classification is a data classification engine that automatically identifies and labels sensitive data across file shares, databases, applications, and data streams. The engine scans structured databases (SQL, Oracle, PostgreSQL), unstructured files (PDF, DOCX, XLSX), semi-structured formats (JSON, XML, CSV), images, and streaming data. It performs full-coverage scans — scanning 100% of rows and files rather than sampling — to reduce edge-case mislabels. Classification goes beyond regex-based pattern matching by combining pattern matching with contextual analysis, named-entity recognition, and AI-assisted methods. Each detected data item is assigned a confidence score, and policies can define thresholds to control which findings are reported. Multi-step validity scoring applies algorithmic checks (e.g., Luhn algorithm for payment card numbers, Soundex for names) to reduce false positives. The system includes cross-field and schema awareness, correlating values with column types, neighboring fields, and known data models to suppress coincidental matches. An operator feedback loop allows confirmed false positives to be recorded so future scans skip the same patterns or locations. Custom data handlers and classifiers can be defined to address proprietary data types or specific governance and compliance requirements. Classification results can be used to enforce data protections such as masking, tokenization, or encryption, and to support regulatory compliance and audit scope reduction. A documented use case involves using the product to support Data Subject Access Requests (DSARs) and "Right to be Forgotten" requests, with findings persisted in a graph database (GraphDB) to enable pivoting and reporting across connected systems.