An Integrated AI-Based Forensic Analysis Platform for Multimodal Biometric, Genetic, and Digital Evidence

Abstract

Forensic investigations increasingly combine visual, biometric, genetic, and digital evidence, yet the analytical tools used for these modalities are commonly separated from case management and evidence handling. This study presents a modular forensic analysis platform that integrates four AI-assisted functions - deepfake screening, face recognition, SNP-based phenotype analysis, and face restoration with downstream demographic estimation - within a case-centric web and mobile workflow. The scientific contribution is architectural and workflow-level: established analytical models are orchestrated through a common evidence lifecycle so that an authorized practitioner can register a case, attach evidence, invoke the appropriate analytical service, and retain the resulting output with the case record. The reported internal evaluation is intentionally module-specific. The deepfake classifier achieved 99.52% accuracy on a balanced 20,000-image test set (19,904 correct predictions); the five-identity face-recognition integration test achieved 94.55% accuracy (52/55); Gradient Boosting produced the highest reported accuracy in the encoded SNP-classification benchmark at 98.34%; and the downstream demographic stage achieved 87% accuracy for the dataset's gender label and 56% for age-group classification. Face restoration is demonstrated qualitatively because no paired-reference restoration metric was recorded in the available experiment. These results are not treated as directly comparable across tasks or as evidence of universal forensic validity. Instead, they show that heterogeneous analytical modules can be combined into one controlled workflow while preserving task-specific interpretation. The platform is positioned as a decision-support environment in which algorithmic outputs require professional review rather than as an autonomous evidentiary decision maker.