Enhancing Plant Disease Classification through Manual CNN Hyperparameter Tuning

Abstract

Over the past few years, the agricultural industry has experienced a significant shift driven by technological progress, increasing environmental awareness, and changing demographics within the farming community. Smart agriculture innovatively utilizes digital technology, real-time data analysis, and precise resource management to promote environmentally responsible farming, employing integrated sensor networks and precision farming techniques for monitoring plant health in both soil and soilless agriculture. Despite the promising benefits, there are significant challenges in ensuring security and privacy of data and the resilience of smart agriculture systems, which are equally applicable to both soil and soilless agriculture. Previous research efforts have explored mechanisms, but these solutions often encounter issues related to authentication, confidentiality, data privacy, and vulnerability to known attacks. Furthermore, current schemes exhibit performance deficiencies due to an excessive reliance on Rivest-Shamir-Adleman (RSA), bilinear pairing, and elliptic curve cryptography (ECC). To overcome all these issues, we introduce a novel method that utilizes homomorphic signcryption based on hyper-elliptic curve cryptography (HECC), addressing the security concerns in both soil and soilless agriculture. This cryptographic technique enhances security while reducing computational and communication demands in comparison to existing mechanisms. Performance analysis indicates that this method is more efficient and offers data privacy, data integrity, confidentiality, authentication, and resistance against known attacks. To validate its safety and robustness, simulation using the Scyther security validation tool were conducted, confirming its appropriateness for smart agriculture environments based on soil and soilless agriculture.