A Delay Tolerant Network (DTN) can effectively operate all network functions and is specifically designed for vast distances under extreme conditions, such as space communications. Traditional Internet cannot perform well in such networks due to high latency, frequent disruptions, and elevated error rates. These issues can be mitigated by a DTN, which provides reliable data transmission and ensures dependable internetworking. Previous research has focused on terrestrial networking methodologies and demonstrated that they cannot resolve the existing issues and challenges of space communication applications; consequently, researchers have shifted their attention towards DTNs. The Space Data System Committee has proposed several protocols to incorporate advancements in DTN. However, their efficiency does not surpass that of DTN, and they lack an automatic data transmission protocol. The current protocols are not designed to transmit data quickly, and the entire transmission process fails if a single node malfunctions or behaves improperly in routing. Therefore, it is crucial to design and implement a novel protocol for DTN that can identify and address misbehaving nodes involved in data transmission based on their identity and location. In this paper, we devise an LSTM-based approach to enhance DTN performance by detecting missing nodes and accelerating data transmission speed. The simulation is conducted using NS2 software, and the results are evaluated concerning delay, packet delivery ratio, abnormal node detection rate, and throughput. The performance of the proposed protocol is compared to that of a similar one to assess its effectiveness.