One of the worst types of cancer is skin cancer. Unrepaired deoxyribonucleic acid (DNA) in skin cells results in genetic errors or mutations on the skin, which is the cause of skin cancer. Skin cancer is best diagnosed early because it is more treatable in its early stages and tends to spread gradually to other body areas. Early diagnosis of skin cancer signs is imperative due to the disease's rising incidence, high death rate, and cost of care. The accuracy of traditional skin cancer diagnostic techniques, especially those that depend on visual examinations, is limited and not accurate, which could endanger the patient. As a result, the use of Deep Learning (DL) has aided researchers in creating a variety of early detection methods for skin cancer. These methods employed characteristics of the lesion, such as color, size, shape, symmetry, etc., to identify skin cancer and differentiate it from melanoma. This paper proposes a new DL-based skin cancer detection and diagnosis scheme (DL-SCDDS) that uses the Human Against Machine 10,000 (HAM10000) dataset, a large and diverse dataset, to ensure an accurate yet effective diagnosis through the implementation of DL techniques, specifically Convolutional Neural Networks (CNN). Before testing, the suggested CNN model underwent training, and it achieved remarkable results, accurately diagnosing seven different types of skin lesions with 96.9% accuracy. Additionally, the results obtained were contrasted with those of other studies that suggested a slightly different methodology; in these comparisons, the suggested model proved to be superior.