The goal of this research is to develop an advanced intelligent model capable of identifying spam calls and fake news, while also enhancing the reliability and accuracy of spam detection. Available models face numerous challenges, including low accuracy, high error rates, slow processing speeds, and difficulties in handling noisy, unbalanced, and complex datasets. To overcome these disadvantages, the current work presents a new model, known as the Arctic Puffin-based Convolutional Detection Model (APbCDM). This model incorporates suitable preprocessing methods with the Arctic Puffin Optimization (APO) algorithm to extract meaningful features and suppress noise in the dataset. It then applies the filtered features through a convolutional network for appropriate classification. The results of the experiment indicate that the model proposed in this paper achieves an accuracy level of 99 percent. It also has precision, recall, and F1-score levels of 99 percent, along with a low error rate of just 1 percent, clearly indicating its superiority over traditional methods. The current work presents a robust, precise, and intelligent approach to detecting spam calls and fake news. Index Terms—Fake News, Convolutional Neural Network, Data Preprocessing, Feature Selection, Spam Call Detection.