The primary mode of COVID-19 transmission is through respiratory droplets that are produced when an infected person talks, coughs, or sneezes. To avoid the fast spread of the virus, the WHO has instructed people to use face masks in crowded and public areas. This paper proposes the rapid real-time face mask detection system or RRFMDS, an automated computer-aided system to detect a violation of a face mask in real-time video. In the proposed system, single-shot multi-box detector is utilized for face detection, while fine-tuned MobileNetV2 is used for face mask classification. The system is lightweight (low resource requirement) and can be merged with pre-installed CCTV cameras to detect face mask violation. The system is trained on a custom dataset which consists of 14,535 images, of which 5000 belong to incorrect masks, 4789 to with masks, and 4746 to without masks. The primary purpose of creating such a dataset was to develop a face mask detection system that can detect almost all types of face masks with different orientations. The system can detect all three classes (incorrect masks, with mask and without mask faces) with an average accuracy of 99.15% and 97.81%, respectively, on training and testing data. The system, on average, takes 0.14201142 s to process a single frame, including detecting the faces from the video, processing a frame and classification.

The face mask detection system has been a valuable tool to combat COVID-19 by preventing its rapid transmission. This article demonstrated that the present deep learning-based face mask detection systems are vulnerable to adversarial attacks. We proposed a framework for a robust face mask detection system that is resistant to adversarial attacks. We first developed a face mask detection system by fine-tuning the MobileNetv2 model and training it on the custom-built dataset. The model performed exceptionally well, achieving 95.83% of accuracy on test data. Then, the model’s performance is assessed using adversarial images calculated by the fast gradient sign method (FGSM). The FGSM attack reduced the model’s classification accuracy from 95.83% to 14.53%, indicating that the adversarial attack on the proposed model severely damaged its performance. Finally, we illustrated that the proposed robust framework enhanced the model's resistance to adversarial attacks. Although there was a notable drop in the accuracy of the robust model on unseen clean data from 95.83% to 92.79%, the model performed exceptionally well, improving the accuracy from 14.53% to 92% on adversarial data. We expect our research to heighten awareness of adversarial attacks on COVID-19 monitoring systems and inspire others to protect healthcare systems from similar attacks.