Runtime And Security Evaluation Of Lightweight Biometric Models For Edge Based Access Control

Abstract

This Study Presents A Runtime And Security Evaluation Of Lightweight Biometric Models For Edge-based Access Control. The Work Evaluates MobileNetV2 For Face Recognition And EfficientNetB0 For Fingerprint Recognition Within An Edge-aware Access-control Environment. The Models Were Assessed Using Locally Captured Biometric Datasets And Public Benchmark Datasets, With Emphasis On Verification Accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR), Equal Error Rate (EER), ROC-AUC, Model Size, Inference Latency, Throughput, And Real-time Decision Performance. The Results Show That The Lightweight Biometric Models Can Provide Reliable Identity Verification While Maintaining Practical Runtime Performance For Edge Deployment. The Face Model Achieved Strong Verification Performance With Low Error Rates, While The Fingerprint Model Provided Complementary Biometric Evidence For Improved Access Reliability. The Findings Demonstrate That Lightweight Deep Learning Models Can Support Secure, Fast, And Scalable Biometric Access Control In Edge Computing Environments Where Low Latency, Privacy Protection, And Real-time Authentication Are Required.

Keywords