Recognition of Valid QR Codes with Machine Learning

The widespread use of Quick Response (QR) codes has made QR codes an attractive target for cyberattacks, posing a security and privacy concern. Quick Response (QR) codes have revolutionized marketing strategies by providing a fast and efficient way to connect offline and online platforms. However, the security issues arise as the usage of QR codes. This paper introduces an automatic detector that identifies QR code images and lets the user know whether it is malicious or benign. To achieve this, a dataset of 10,000 original and malicious QR code images (both) is taken. Convolutional Neural Network (CNN) is employed to classify original and malicious codes. Additionally, histogram density analysis is incorporated to enhance feature identification, leading to improved classification accuracy. This paper implemented ResNet50, MobilenetV3 and InceptionV3 CNN models for evaluating the abilities to predict the malicious codes. The InceptionV3 model emerges as the most promising, attaining an accuracy of 98.13%, offering an efficient and accurate solution for document forensics and counterfeit detection.