Enhancing Frontier Security: A Unified Cross-Modal AI Framework for Modern Migration Governance
TL;DR
Results indicate that combining different AI-based checks can support faster and more consistent screening, although further testing with larger and more diverse datasets is required before considering real-world deployment.
Abstract
Immigration authorities need to process many travellers while maintaining adequate security checks. Manual verification of documents and passenger information can become time-consuming, particularly at busy border checkpoints. This study presents a multi-modal AI-based framework that combines three components: sentiment analysis of passenger interactions, OCR-based visa verification, and weight-based anomaly detection. The sentiment analysis module examines textual or conversational cues to identify unusual emotional patterns that may require additional attention. The document verification module uses OCR and computer vision to extract visa information and identify possible inconsistencies. The third module compares passenger weight measurements with available reference information to identify unusual variations that may warrant further inspection. The three modules are combined to provide an overall risk assessment and support immigration officers during screening. In the reported evaluation, sentiment analysis achieved an accuracy of 97.35%, while the OCR-based visa verification and weight-based detection modules achieved 96.42% and 93.75%, respectively. The integrated system recorded an overall accuracy of 95.84%, with an average processing time of 1.76 seconds. These results indicate that combining different AI-based checks can support faster and more consistent screening, although further testing with larger and more diverse datasets is required before considering real-world deployment.