With the advent of hyper-connected systems and dynamic cyber threats, secure and efficient data transmission within the contemporary communication networks has become paramount of priority. The solutions that are offered by this paper are a hybrid intelligent scheme that comprises Deep Q-Network (DQN), a deep reinforcement learning method, and NTRU, a lattice-based public key encryption algorithm, to strengthen data security in dynamic network scenarios, including those of Mobile Ad Hoc Networks (MANETs) and the Internet of Things (IoT). The main idea is that the suggested model uses an adaptive learning ability of the DQN to decide in real-time regarding routing and encryption schemes depending on network conditions, such as node density, mobility, the stability of links, and traffic characteristics. At the same time, NTRU provides post-quantum security, efficient encryption and decryption operations, which is applicable to the resource-restrained devices. In order to confirm the designed framework, NS2 extensive simulations were carried out by employing different performance metrics that include throughput, energy consumption, packet delivery ratio and end to end delay. Experiments indicate that the proposed DQN-NTRU hybrid model is more efficient and secure than the conventional encryption algorithm, such as RSA and ECC, in terms of security strength and computation alacrity.