Skin cancer is one of the most commonly diagnosed types of cancer, affecting the health of millions of people each year. The survival rate of skin cancer is highly dependent on the stage at which cancer is detected. Survival rate is comparatively higher when it is detected in the early stage. However, skin cancer detection is complex and expensive procedure, hence it becomes difficult to detect it in early stage. In this research, we proposed a deep learning-based algorithm for detection and identification of skin lesion as malignant or benign based on the image captured. Data preprocessing is used for segmentation and removing unwanted features and noise. Neural network algorithm is used to train the model for classification of lesions. In this research, dataset is collected from internet sources and greyscale image conversion, homomorphic filtering and segmentation are applied. To generate the model, Artificial Neural Network with Bayesian Regularization is used. Artificial neural network tunes the weights by reducing the error function, but Bayesian Regularization uses probability distribution of network weights, which increases the accuracy of system. For the performance analysis, different datasets are used as an input and performance is evaluated by comparing the result with existing algorithms.