The rapid adoption of Internet of Things (IoT) technologies in precision agriculture has significantly enhanced farm productivity, resource efficiency, and real-time decision-making. However, this digital transformation also exposes agricultural infrastructures to a growing range of cyber threats, including network intrusion, data manipulation, and service disruption. Traditional Intrusion Detection Systems (IDS) often fall short in these environments due to their high computational complexity and limited adaptability to flow-based traffic patterns inherent in agricultural IoT systems. In this paper, we propose Farm-FlowNet, a lightweight Convolutional Neural Network (CNN) based IDS tailored for flow-based data in precision agriculture networks. The model is designed to learn and classify traffic patterns directly from the Farm-Flow dataset, an emerging benchmark featuring labeled attack types and normal flow records specific to AG-IoT systems. Farm-FlowNet employs a compact 1D CNN architecture optimized for edge deployment, ensuring both computational efficiency and high detection performance. Extensive experiments demonstrate that the proposed model achieves competitive accuracy, precision, and recall while maintaining low inference latency, making it well-suited for real-time deployment on resource-constrained agricultural nodes. Additionally, the use of a flow-based representation enhances the model’s ability to detect subtle attack signatures without relying on packet-level inspection. The model’s performance is evaluated using standard metrics, and results are benchmarked against conventional machine learning classifiers. Farm-FlowNet marks a step toward secure, efficient, and interpretable IDS design for next-generation smart farming ecosystems. Future extensions include integrating explainability modules and federated learning for scalable, privacy-preserving threat detection.