Delay tolerant network is a boon in emergency fields like flood and war zones. The data gathered by the sensor nodes is transmitted whenever any aggregator node comes in contact with those stationary sensor nodes. However, few nodes can behave differently and don't transmit the information. These nodes which don't take part in communication to preserve it's battery or are compromised are selfish nodes and have to be identified to avoid communication disruption. This article discusses the selfish node's detection schemes and proposes a novel hybrid scheme. Most of the recent schemes work with on node's reputation or incentives if it takes part in communication. This paper has proposed an incentivized reputation scheme that first clusters the nodes using their social features, calculates the weighted social tie as their social connection strength and updates the weighted social tie by applying reward or penalty. The incentive is offered if residual energy and packet delay has a tradeoff or are penalized. A new modularized deep nonnegative matrix deep autoencoder is developed to calculate the reputation of nodes using social features and named IRU-mDANMF (incentivized reputation update by modularized DANMF). The scheme has been experimented with in several scenarios and is performing significantly better than state-of-the-art schemes.

Summary A sensor node carries out a specific function in a wireless sensor network. Wireless sensor networks are less bearable but more vital to the industry and the people because of threats. As a result, a node's battery life will be drastically reduced, and the node will be rendered completely inoperable, which is the most extreme form of denial of service attack. A vampire attack, one of the denials of service attacks, may inflict extensive harm to the network, making it harder to detect and using more energy than necessary. This article has proposed a novel vampire attack detection and prevention by the energy consumption prediction in the data path with the least error as low as . In combination with the social spider optimized Gaussian mixture model, the gray prediction model is used to detect and prevent the attack. The energy prediction scheme calculates a cooperative trust score and is categorized by the optimized Gaussian mixture model. The algorithm is validated with recent state‐of‐the‐art schemes and the detection accuracy improvement of up to 35.27% is achieved.

Abstract Sensor nodes in WSN play a vital role in communication, IoTs and many other emergencies too. However, the energy consumption of nodes is a major setback to these, which incites various malicious nodes/attacks. This article studies and presents the solution to Vampire attack‐ one of those kinds of attacks. It depletes the energy by route elongation of data transmission. This article has suggested a novel two‐fold mechanism to detect the attack by integrating co‐operation trust mechanism and the mitigation of the attack by selecting the secure route by policy gradient‐deep reinforcement learning. The designed protocol also ensures the selection of a secure hop even in the presence of the vampire attack. The results are compared with various other existing state‐of‐the‐art schemes and have improved the detection ratio by 20% compared to the forecasting methods applied for detecting the vampire node's behavior. The network lifetime has also improved by 3% than the benchmark dynamic source routing.