Steganography is the art of hiding sensitive data in a document or image without changing the content itself. On the other hand, steganalysis is the process of identifying and examining information that is concealed under a cover media. A deep learning-based framework for the steganalysis of LSB-encoded images is proposed in this study. The framework consists of a Convolutional Neural Network (CNN) with an L2 regularization model trained to distinguish between stego and cover images. 2500 cover and 2500 stego photos, divided into training and testing sets, make up the dataset used in this investigation. Data augmentation strategies are used to increase the size of the training set. The proposed CNN model surpasses the accuracy found by previous studies with a test accuracy of 95%. The accuracy, precision, recall, and F1-score are among the measures used to assess the model’s performance. A thorough evaluation of the model’s performance is provided by printing the classification report and confusion matrix. To see the model’s accuracy throughout training and validation, the training history is plotted. The suggested methodology can be used to identify and stop malware and other cyberthreats that conceal their dangerous payloads using steganography in a variety of domains, including cybersecurity, law enforcement, and digital forensics.