Cyberbullying is the hurtful harming or harassment of people through the use of social media platforms including messaging apps, blogs, e-commerce, and websites. This kind of activity can imply posting troublesome content about someone, threatening comments, or outspread rumors about an individual. The act of hurting or assaulting someone online, or cyberbullying or cyberstalking, has serious emotional implications. To recognize the cyberbullying in Assamese Language on social media we used two Deep Learning techniques, i.e., Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU). These techniques successfully differentiated between cyberbullying related to gender, religion, age, ethnicity, and non-cyberbullying content using a balanced dataset consisting of 39,869 social media comments. The CNN model outperformed the BiGRU model by reaching an accuracy of 90.3%. These results show the effective detection of cyberbullying in the Assamese Language paving the path for a more safer and secure online environment.