Identifying Sensors Data Integrity Threats of Smart Agriculture: A Collaborative Filtering Approach

Abstract

Listen

Translate article

HighlightsData integrity threat is one of the challenges faced by smart agriculture, where humans make complex decisions based on the results and recommendations of automated farm/livestock monitoring systems.Smart or digital farming heavily relies on data, which can face various integrity threats including data modification, data destruction, etc.The use of collaborative filtering as a lightweight method to detect abnormal data has proven effective from our experimental study. It can detect suspicious data with high accuracy and precision enhancing digital agricultural security.Abstract. Smart agriculture leverages human intelligence and artificial intelligence in digital farming, which heavily relies on data collection, analysis, and decision support delivery. Farmers need reliable data and a decision support system to make accurate decisions and take appropriate actions. The Internet of Things (IoT) not only creates the backbone for farm data collection for intelligent farming but also brings some security concerns to the agriculture field. Though many studies exist on security in agriculture, particularly device identification and tracking, cryptography, blockchain, and other security approaches, the concerns about agricultural data integrity have not been fully explored in the current literature. Data integrity threats could cause huge losses to farmers, threatening food security, especially in developing countries. Given the early stage of development in the domain and the rapid adoption of IoT, ensuring the trustworthiness of data from various devices poses a significant challenge. This challenge is exacerbated in wireless sensor network scenarios, particularly in harsh environments where potential avenues for physical attacks exist. This article evaluates a data integrity threat detection technique in an IoT wireless network. It proposes an approach to identify potential data integrity failures or threats. The effectiveness of this approach is demonstrated through an empirical use-case study focusing on agriculture applications. Through experimental and trace-based simulations, we illustrate that threats can potentially be identified with a 91% accuracy rate and approximately 98% precision and recall. The proposed solution could be deployed in distributed and centralized digital agricultural systems to identify and predict real-time data integrity issues or threats. Keywords: Collaborative filtering, Data integrity threats, Local outlier factor, Smart agriculture.