Abstract—Mobile Ad Hoc Networks (MANETs) are potent technologies frequently used to collect data from their environments and disseminate it to other nodes or servers. Many cyber threats such as Denial of Service (DoS) attacks and black hole attacks are exposed to these networks because messages are exchanged over open public networks. These networks are vulnerable to attacks due to their decentralized and dynamic nature. In black hole attacks adversaries conceal their presence by redirecting traffic to incorrect points and dropping data packets. In this work a deep learning-based SA-DCBiGNet stands for Self-Attention– Dense Convolutional Bidirectional Gated Network with a twin attention mechanism is presented to protect MANETs as dynamic networks from blackhole attacks by detecting them. The performance of the model is tested in a python platform using the dataset of the WSN-DS IoT. The findings indicate that this solution incurs accuracy, precision, F1−Score, recall and of 99.70%, 99.00%, 99.02% and 98.3% respectively.