Abstract In Deep learning contains procedures that are initially trained in context to “learn” their functions and then used with invisible input for sorting dedications. The benefits of Mobile Adhoc Network (MANETS) and the growing demand have attracted a percentage of attention from the research community. However, it appears to be extra vulnerable to several attacks affecting presentation than other types of networks. The Intrusion Detection System (IDS)delivers a second line of protection against manipulation by monitoring network activity to investigate malicious attempts by attackers. Due to Manet underlying distributed architecture, traditional cryptographic systems cannot fully protect Manet from new threats and insecurities. Implementing in-depth technology for IDS can meet these challenges. In this paper, the simulation stage was developed with a NS2 simulation platform. The RNN classification algorithm was evaluated using several metrics to detect intruders. The efficiency of RNN as an approximate tool for detecting, isolating, and reconfiguring attacks was measured on datasets with different data traffic situations and dynamic patterns for manifold attacks. With a final search rate of 0.32 to 2. 35%, this feature not only provided a creative and less effective way to carry out man-in-the-middle attack (MITM) attacks on modelstages, but also made it an important factor in identifying and isolating such attacks. In addition to existing IDs, this work is intended for future generation, identification, isolation, and redesign of malicious software.