Malware is software utilized for undergoing various malicious activities, including the theft of money and passwords. The Internet of Things (IoT) devices are affected by various losses because of dominant malware variants. In addition, malware is harmful in several ways, such as stealing, deleting, or encrypting data, and a system becomes unusable or locked. Malware variant traffic identification is done by malware detectors by various anti-malware methods, such as virus signature scanners and heuristic methods. However, they suffer from several limitations including computational complexity, large number of training data, and vulnerability. Thus, Quantum Dilated Convolutional Neural Network (QDCNN) with Wild Horse Optimization (WHO) is introduced to address these problems. At first, a system model of IoT Edge Gateway is considered, and then, log file collection is done. After that, features like network-oriented features and traffic-oriented features are mined from the collected log file. Later, the extracted features are forwarded to an augmentation process, where oversampling is exploited to address the class imbalance problem. At last, malware variant traffic identification is effectuated by utilizing the proposed WHO_QDCNN. Moreover, the developed WHO_QDCNN attained better precision, accuracy, and recall of 93.77%, 90.05%, and 94.66%, respectively.