The current advancement in Unmanned Aerial Vehicles (UAVs) and the proliferation of the Internet of Things (IoT) devices is revolutionizing conventional farming operations into precision agriculture. The agricultural UAVs combined with IoT use an open channel i.e., the Internet to assist cultivators with data collection, processing, monitoring, and making correct decisions on the farm. However, the use of the Internet opens up a wide range of challenges such as security (e.g., performing cyber-attacks), risk of data privacy (e.g., data poisoning and inference attacks), etc. The usage of current conventional centralized security measures has limitations in terms of a single point of failure, verifiability, traceability, and scalability. Motivated from the aforementioned challenges, we propose a Secured Privacy-Preserving Framework (SP2F) for smart agricultural UAVs. The proposed SP2F framework has two main engines, a two-level privacy engine, and a deep learning-based anomaly detection engine. In the two-level privacy engine, a blockchain, and smart contract-based enhanced Proof of Work (ePoW) is designed for data authentication, and to mitigate data poisoning attacks. A Sparse AutoEncoder (SAE) is applied for transforming data into a new encoded format for preventing inference attacks. In the anomaly detection engine, a Stacked Long-Short-Term Memory (SLSTM) is used to train and evaluate the results of the proposed two-level privacy engine using two publicly accessible IoT-based datasets, namely ToN-IoT and IoT Botnet. Finally, based on thorough analysis, and comparison, we identify that the SP2F framework outperforms several state-of-the-art techniques in both non-blockchain and blockchain frameworks.

当前无人飞行器（UAV）的发展以及物联网（IoT）设备的普及，正在将传统的耕作方式彻底变革为精准农业。农业无人机与物联网相结合，使用开放渠道，即互联网，以帮助中耕者在农场上收集数据，处理，监控并做出正确的决策。但是，Internet的使用带来了广泛的挑战，例如安全性（例如，执行网络攻击），数据隐私的风险（例如，数据中毒和推断攻击）等。当前常规集中式安全措施的使用在单点故障，可验证性，可追溯性和可伸缩性方面有局限性。出于上述挑战，我们提出了一种用于智能农业无人机的安全隐私保护框架（SP2F）。提出的SP2F框架具有两个主要引擎，一个两级隐私引擎和一个基于深度学习的异常检测引擎。在两级隐私引擎中，区块链和基于智能合约的增强型工作量证明（ePoW）设计用于数据认证，并减轻数据中毒攻击。稀疏自动编码器（SAE）用于将数据转换为新的编码格式，以防止推理攻击。在异常检测引擎中，使用堆栈式长短期内存（SLSTM）使用两个可公开访问的基于IoT的数据集，即ToN-IoT和IoT僵尸网络，来训练和评估建议的两级隐私引擎的结果。最后，基于全面的分析和比较，