A Distributive Correlated Neural Network (DCNN) Approach for Clone Node Identification in Hybrid WSNs

In this research article, a novel approach using the Distributive Correlated Neural Network (DCNN) algorithm is developed and presented along with simulation results. Wireless sensor networks (WSNs) are composed of spatially dispersed sensor nodes designed to observe and record environmental or physical data. These nodes, typically battery-powered and mobile, often struggle to meet the objectives of extended operational life and high dependability. While static, energy-harvesting nodes offer longer lifespans by converting ambient energy into electrical energy, they tend to be costlier. A hybrid system, integrating both mobile and static nodes, can help balance these conflicting requirements of durability and cost-efficiency. Clone attacks pose a significant risk to such hybrid WSNs. These attacks are feasible due to the ease with which adversaries can extract configuration and authentication data from non-tamper-proof nodes and duplicate them within the network. This study introduces an innovative clone detection method that is particularly suitable for hybrid WSNs because it: (1) does not require location data of nodes, (2) functions independently of the network topology, and (3) supports hybrid architectures comprising both static and mobile nodes. Additionally, the detection algorithm's design allows for parallel execution, significantly speeding up the overall detection process. In modern applications, sensors are often mounted on mobile platforms or devices to monitor operational parameters and ensure proper functionality. These systems rely on WSNs for real-time monitoring and control via remote servers. Data is stored and managed securely on cloud-based systems, yet the use of publicly accessible networks during transmission introduces vulnerabilities. Malicious actors can intercept and manipulate sensor data, which may lead to corrupted databases and misuse of sensitive information provided by healthcare professionals or technicians. This work addresses these security concerns by developing a system capable of predicting and identifying cloning-based attacks. The model uses a classification approach that analyzes data from IoT-enabled sensors to distinguish between normal and cloned devices. When a device is classified as normal, data transmission proceeds as usual. However, if a cloning attempt is detected, it is flagged and sent to a firewall or routing mechanism to block further transmission. Concurrently, the system updates the classification model by learning new features of the identified clone, thereby enhancing the model’s ability to recognize future threats. To optimize cluster formation of sensor nodes, the CREN technique is utilized in conjunction with the DCNN classification model. This approach facilitates the grouping of sensor data based on multiple parameter combinations to improve predictive performance. Key parameters used for feature evaluation include Received Signal Strength Indicator (RSSI), entropy, sensor output weights, energy metrics, trust scores, Lebesgue measures, sequence probability densities, overall power consumption, packet transmission count, likelihood ratios, and average sigma values. These features are crucial for determining clone likelihood prior to transmission. The system segments the dataset into clusters based on the MCC-derived clustering index and assesses the correlation between training and test data. This optimal clustering and feature classification process enhances detection performance and accuracy. Evaluation metrics such as sensitivity, specificity, precision, recall, and accuracy—measured against ground truth—are employed to validate the effectiveness of the proposed method.