<![CDATA[Transforming the digital payment system through Quick Response Code Indonesian Standard (QRIS) Static Merchant Presented Mode (MPM) has provided convenience for MSMEs, yet also triggered significant security risks, particularly quishing attacks involving fraudulent sticker overlays. This research aims to develop a comprehensively integrated fraud detection system using the YOLOv11 Single-Stage Detector architecture and a rule-based inference engine. The research methodology includes the construction of a representative dataset of 898 images and the implementation of a three-layer validation mechanism comprising spatial layout analysis based on ASPI standards, textual semantic validation using PaddleOCR, and geospatial verification of GPS coordinates. The system utilizes the SequenceMatcher algorithm with adaptive thresholds to accommodate merchant typography variations. Experimental results indicate that the YOLOv11-m variant provides the best localization accuracy with a mean Average Precision (mAP) 50-95 score of 0.8682. End-to-end evaluation test images yielded an overall system accuracy of 96.15%. Significantly, the system achieved a recall of 1.00 for the suspicious class, proving its ability to identify all potential visual manipulation threats without omission. Although blurred images lowered the authentic class recall to 0.90, security principles remained intact by classifying unvalidated data as suspicious. This study provides a significant contribution to strengthening digital payment integrity through a precise, lightweight detection mechanism aligned with Indonesian national regulations.]]>
