Lamport Certificateless Signcryption Deep Neural Networks for Data Aggregation Security in WSN

Abstract

Confidentiality and data integrity are essential paradigms in data aggregation owing to the various cyberattacks in wireless sensor networks (WSNs). This study proposes a novel technique named Lamport certificateless signcryption-based shift-invariant connectionist artificial deep neural networks (LCS-SICADNN) by using artificial deep neural networks to develop the data aggregation security model. This model utilises the input layer with several sensor nodes, four hidden layers to overcome different attacks (data injection, compromised node, Sybil and black hole attacks) and the output layer to analyse the given input. The Lamport one-time certificateless signcryption technique involving three different processes (key generation, signcryption and unsigncryption) is adopted to achieve secure data transmission between the sender and receiver. Firstly, a one-way function is executed to generate the public and private keys for each sensor node in the WSN. Secondly, digital signature generation and encryption are both performed. The sender node, which handles the signature generation and data encryption, forwards the data to the aggregator node. Then, the receiver verifies the data by using the sender’s public key during data decryption. Thus, data aggregation security can be guaranteed. Finally, the authorised node aggregates the data with much higher data confidentiality and privacy. Performance analysis is conducted by simulating the proposed LCS-SICADNN and conventional models. Results of comparison indicate that the LCS-SICADNN can improve data aggregation security with higher throughput and lesser delay, packet drop and overhead compared with the conventional methods.