A secure and energy-efficient edge computing improved SZ 2.1 hybrid algorithm for handling iot data stream

IoT devices generate a massive amount of sensitive data, which is transferred tremendously to the cloud for processing and decision-making. The most significant issues that need to be solved for IoT devices are improving energy efficiency and guaranteeing data security while several devices are connected. In this paper, for edge computing, a hybrid algorithm is proposed that uses compression and encryption in the same manner to improve efficiency in terms of energy and data security in IoT devices. The authenticated encryption with associated data (AEAD) ChaCha12-Poly1305 algorithm and improved SZ 2.1 compression are used in this hybrid architecture. While sending the data to the edge, confidentiality, integrity, and authentication are maintained. Several experiments were conducted considering the driver stress detection dataset (Bernstein DJ. Lecture Notes in Computer Science including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics 4986:84–97 (2008); Lakshminarasimhan S et al. LNCS 6852(1):366–379 (2011)) and gas-sensor dataset (Ibarria L et al. Comput Graph Forum 22(3):343–348 (2003)) for monitoring the home activity to validate this approach. The performance of the proposed improved SZ 2.1 compression algorithm is compared with the five baseline algorithms including SZ 1.4, original SZ 2.1, selective compression, algorithm-based fault tolerance (ABFT), and digit rounding algorithms. The key parameters used in the experiment to measure the performance of the proposed algorithm are data reduction, compression ratio, power consumption, encryption time, total processing time, and error. Using the improved SZ 2.1 compression algorithm in conjunction with the ChaCha12-Poly1305 AEAD algorithm, the device’s battery life is also enhanced by about 10% while ensuring the security of data disseminated to the Edge. The use of the proposed secure hybrid model reduces both encryption and overall processing time by 95% and 98% respectively for both datasets.