A privacy-protecting eggplant disease detection framework based on the YOLOv11n-12D model.

Abstract

The growing global population and rising concerns about food security highlight the critical need for intelligent agriculture. Among various technologies, plant disease detection is vital but faces challenges in balancing data privacy and model accuracy. To address this, we propose a novel privacy-preserving eggplant disease detection system with high accuracy. First, we introduce a lightweight 3D chaotic cube-based image encryption method that ensures security with low computational cost. Second, a streamlined YOLOv11n-12D framework is employed to optimize detection performance on resource-constrained devices. Finally, the encryption and detection modules are integrated into a real-time, secure, and accurate identification system.Experimental results show our framework achieves near-ideal encryption security (entropy=7.6195, Number of Pixel Change Rate(NPCR)=99.63%, Unified Average Changing Intensity(UACI)=32.85%) with 23× faster encryption (0.0127s) versus existing methods. The distilled YOLOv11n-12D model maintains teacher-level accuracy (mAP@0.5=0.849) at 3.6× the speed of YOLOv12s (2.7ms/inference), with +6.5% mAP improvement for small disease detection (e.g., thrips). This system balances privacy and real-time performance for smart agriculture applications.

Code and data availability

The supplied article blocks describe a private eggplant disease image dataset (8,265 images) and a YOLOv11n-12D detection/encryption framework, but contain no data availability statement, no public dataset deposit, and no author code repository or URL. No qualifying paper-specific public assets are present.

No evidence-backed public reproduction asset is currently recorded.