Anomaly detection in smart agriculture systems

This work proposes a new approach for Anomaly Detection in smart agriculture systems. Through the use of multi-sensor systems and Decision Support Systems, it is possible to collect, analyze, and process huge amounts of data on agriculture. This supports the farmer in the decision-making process to optimize the results in terms of quality and quantity, to avoid waste, and to maximize profits. However, the use of IoT and intelligent communication technologies can introduce a number of intentional and unintentional weaknesses and flaws in data and information management. The research proposes an Anomaly Detection System (a multi-layered architecture) to mitigate the infrastructure threats in a pro-active way in the Smart Agriculture domain. The design of the proposed architecture is based on a machine-learning algorithmic approach by a multivariate linear regression (MLR) and a long-term memory neural network algorithm (LSTM). The application of the Anomaly Detection System was done on a real dataset coming from a smart agriculture system located in the Apulia region (Italy).