False information and misleading content can have disastrous consequences, even posing threats to lives. One particularly pervasive form of this is fake news, which has surged with the growth of the social media. The rapid increase in the spread of misleading information is alarming, as it travels much faster than real news since the source of the fake news cannot be identified, leading to potentially devastating outcomes. Preventing the spread of misleading information is crucial to mitigating these dangers. This paper shows how Ensemble techniques and BERT models can be employed to detect fake news for languages with limited resources, including Tamil, Malayalam, Telugu, and Kannada. The experimentation shows that BERT-base-multilingual-cased and Stacking Techniques reach the highest accuracies of approx. 96% and 90% respectively for all the languages.