Skin cancer, a common and life-threatening condition commonly resulting from erythematous dermatosis, requires early detection to minimize fatality. The conventional methods of diagnosis, mainly founded on clinical knowledge and biopsies, are time-consuming, arbitrary, and susceptible to false results. The current research introduces a novel deep learning method using Convolutional Neural Networks (CNNs) for effective skin cancer identification from erythematous dermatosis. The model is learned using a big data set of skin lesion images to distinguish between benign and malignant diseases. In contrast to traditional techniques, this CNN-based system provides automatic, quick, and highly accurate diagnoses, minimizing the need for invasive tests and enhancing diagnostic efficiency. The precision of the proposed method, its cost-effectiveness, and speed of diagnosis make it a useful tool for dermatologists, facilitating early detection and treatment, and resulting in improved patient outcomes. This study is a necessary step toward the incorporation of deep learning into dermatology, offering a high-performing alternative to conventional methods. Recent research has shown that CNN-based methods are over 91% accurate in detecting skin cancer, proving their potential for broad clinical use.