Detecting Data Poisoning Attacks on Smart Farm Devices Using Machine Learning
Open Access DepositedData poisoning attacks pose significant threats to the integrity of agricultural data. These attacks lead to erroneous decisions and compromised crop production, ultimately affecting food safety, public health, and national security.This praxis uses a comprehensive methodology that includes data collection, exploration, cleaning, feature selection, and rigorous testing of machine learning models. The data collection phase involved obtaining and consolidating 23 comma-separated value files, containing a mix of categorical and numerical data. The focus was on specific attack types relevant to smart agriculture, such as password, backdoor, injection, and man-in-the-middle (MITM) attacks. The research utilized deep neural networks (DNNs) and quantized autoencoders for anomaly detection, leveraging their ability to handle large-scale and imbalanced datasets characteristic of cyber-attack data. The DNN model demonstrated exceptional performance metrics, achieving accuracy, precision, recall, and F1 score values of 99.00%. The model fusion approach using DNNs outperformed traditional single DNN models in terms of F2 score and false alarm rate indicating higher accuracy in detecting true anomalies and reducing false positives. The quantized autoencoder model, optimized for deployment in resource-constrained environments such as Internet of Things (IoT) devices, significantly reduced memory size and peak central processing unit (CPU) utilization while maintaining high performance in detecting anomalies. The research successfully validated machine learning models that can effectively detect data poisoning attacks on smart farm devices. The integration of advanced techniques such as quantization and ensemble learning with DNNs provided robust and efficient solutions for real-time anomaly detection in smart agriculture environments. These findings underscore the importance of employing sophisticated machine learning models to enhance the security and resilience of smart agriculture systems, ensuring the integrity of agricultural data, and safeguarding the global food supply chain.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Mickalson_gwu_0075A_16962.pdf | 2025-04-11 | Open Access |
|