In today’s generation, face detection and recognition plays an important role in applications such as biometrics, surveillance, and human-computer interaction. However, many current methods raise concerns about the safety of stored biometric templates, which could be misused. To tackle this issue, we suggest a privacy-preserving multi-face recognition framework that combines accurate detection, effective feature extraction, and cryptographic protection. The system uses RetinaFace for strong face localization with landmark guidance. For reliable recognition, RetinaFace is used to extract embeddings. The extracted embeddings use SHA-256 hashing technique to protect user identity in non-invertible templates that prevent reconstruction of original biometric data. The dataset shows the framework achieves accuracy 98.3%, precision 96.9%, recall 96.1%, and F1-score 91.8%, demonstrating effectiveness in real-time applications. This result shows outstanding performance in robust recognition while offering a privacy-conscious solution for secure biometric authentication.