Fake news is widely disseminated and constitutes an imminent danger to society since it may shift public opinion, affect elections, and stir up social unrest. For an informed and reliable information ecosystem to continue to exist, the ability to identify fake news in various languages is essential. In this paper, we propose a transformer-based model for Bangla fake news detection using DistilBERT, a distilled variant of the BERT model. We begin by curating a large-scale dataset of Bangla news articles, comprising both authentic and fake news samples. To capture the contextual information and semantic relationships within the Bangla language, we fine-tune the DistilBERT model on this dataset. The pre-training process entails putting the model through its paces on a sizable corpus of Bangla text data. We compare its performance with that of other standard machine learning and deep learning models, such as Support Vector Machine (SVM), Random Forest (RF), and Long Short Term Memory (LSTM) in order to show that our model is superior at correctly identifying fake news in Bangla. The experimental findings show that the suggested model performs competitively in detecting Bangla fake news, surpassing other approaches with 97.85% accuracy.