Experimental Telemetry Dataset and Codebase for Closed-Loop-TinyOL v1

Experimental Dataset & Codebase: TinyOL for Edge Anomaly Detection This repository contains the complete raw hardware telemetry datasets, Jupyter Notebook analysis pipeline, and C++ hardware deployment code from an experimental campaign evaluating the "TinyOL" framework on severely resource-constrained ESP32 microcontrollers. The codebase and data were generated to validate a novel, ultra-lightweight CUSUM-EWMA algorithm featuring an anti-poisoning baseline freeze, designed to protect multi-modal IoT sensor reliability during severe network congestion and DDoS floods. Note: An interactive, live web simulation of this algorithmic framework is publicly accessible via GitHub Pages at: haranathrakshit.github.io/TinyOL-Simulation/. 📄 ASSOCIATED MANUSCRIPT This dataset and codebase support the findings presented in the research article: "Closed-Loop-TinyOL for Autonomic Anomaly Detection, Diagnosis, and Self-Healing in Edge IoT" 📂 REPOSITORY STRUCTURE TinyOL_EdgeML_Codebase/ │ ├── 01_Experiment_Scripts/ (Python attack generators & Arduino firmware) ├── 02_Raw_Datasets/ (Physically validated, 0-NaN hardware telemetry) │ ├── Dataset_1_Stochastic_DDoS.csv │ ├── Dataset_2_Hardware_Saturation.csv │ └── Client Side Arduino Serial Monitor Logs/ ├── 03_Analysis_Notebooks/ (Modular Evaluation Pipeline & Reproducibility) │ ├── 00_Auditing.ipynb │ ├── 01_TinyOL_Detection_Analysis.ipynb │ ├── 02_Resource_Exhaustion_Limits.ipynb │ └── 03_Metrics_Evaluation.ipynb ├── 04_Figures_and_Results/ (High-resolution manuscript plots) └── README.md 📊 KEY DATA / COLUMN DEFINITIONS L_t: Extracted physical MAC-layer transmission latency measured in milliseconds (ms). mu: The EWMA-based baseline mean calculated directly on the silicon. sigma: The EWMA-based mean absolute deviation. S_t: The TinyOL CUSUM Danger Score (accumulation of standardized deviations). State: The anti-poisoning state machine status (Normal, Warning, or Critical). 🛠 USAGE The included Jupyter Notebooks in the 03_Analysis_Notebooks/ folder are numbered sequentially (00 to 03). Executing these notebooks in order will audit the datasets, evaluate the anomaly detection thresholds, and automatically reproduce the classification metrics (e.g., 80.10% Precision) and figures found in the associated manuscript. To replicate the hardware benchmarks and multi-modal sensor fusion, compile and flash the TinyOL_EdgeNode.ino file directly to an ESP32 microcontroller. Environment: Python 3.8+, Scikit-Learn, Pandas, Numpy, Matplotlib, Jupyter, Arduino IDE (v1.8+ or v2.x with ESP32 Core).