The rampant dissemination of false and unsourced information, commonly known as hoaxes, has become a pervasive issue in the era of internet media. In the digital age, the widespread dissemination of false and unverified information has emerged as a critical concern within the realm of internet media. Hoax news can be used to influence elections, sway public opinion, and create political instability. The rapid evolution of information technology has contributed to the uncontrollable proliferation of hoax content, necessitating the development of intelligent systems for effective classification. This research focuses on implementing a robust classification system for identifying hoax news circulating through internet media. The method used in this program is the Naive Bayes method, specifically Naive Bayes Multinomial, which works with the assumption that each feature (word) is considered independent from the others. Text vectorization using CountVectorizer converts text into a numeric vector, which can be used by classification algorithms. This program uses a trained model to make predictions on testing data and calculate evaluation metrics such as accuracy, confusion matrix, and classification reports. By leveraging these methodologies, the study aims to enhance the accuracy and efficiency of distinguishing genuine news from deceptive hoaxes. The highest accuracy value obtained in this research was 94.73% with a division of 20% test data and 80% training data. True Negative (TN): 4555, False Positive (FP): 178 and False Negative (FN): 295, True Positive (TP): 3952