QR codes facilitate seamless information sharing but are increasingly exploited for cyber threats. New assaults are difficult for traditional blacklist-based techniques to identify, hence sophisticated solutions are required. This study proposes a hybrid machine learning (ML) and deep learning (DL) approach to identify and categorize malicious QR codes based on their content. By leveraging ML for classification and DL for feature extraction, our model significantly enhances detection accuracy for text-based, structural, and behavioral data. Our approach outperforms existing methods in detecting known and novel QR code-based threats with greater accuracy, resilience, and flexibility. The proposed method provides an intelligent, effective, and scalable defense against evolving QR code security threats.