Detection of fake news or misleading information on social media is a major concern today. Fake news circulated on social media is multimodal in nature, containing news text, images, videos, and other metadata captured by social media platforms such as the number of tweets, retweets, likes, shares, etc. Existing work mainly uses a single modality for fake news detection for example finding clues from the news text, and detecting photoshopped images, but they fail to provide promising results on the real-time news posts. Considering the multimodal nature of news posts, there is a need to analyze every modality and metadata information and decide the authenticity of the news post. In this paper, we have designed an ensemble-based multimodal system that detects the authenticity of social media news. The system has three different components, which learn the news text, news visual, and social context features. For performing the final classification, an ensemble of the three components is done using average, weighted average, and novel fuzzy reward techniques. The paper undertakes a comparative analysis of three ensemble techniques, discussing the results with various performance metrics including precision, accuracy, F1-score, and the ROC-AUC curve. The ensemble with a weighted average provides a precision of 92.83% and fuzzy rank techniques provide a promising recall of 85.15%.