"TinyML Cybersecurity Dataset for Resource-Constrained Air Quality IoT Monitoring with Freeze and Replay Attack Classification"
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
"This dataset supports the study \"Real-Time On-Device TinyML Cybersecurity for Resource-Constrained Air Quality IoT Monitoring with Freeze and Replay Attack Classification.\" It contains a complete 11-class multivariate air-quality cyberattack dataset for evaluating on-device TinyML detection and classification of sensor-link attacks in a resource-constrained Internet of Things monitoring system.The dataset includes one nominal class and ten attack classes corresponding to freeze and replay attacks on five pollutant channels: PM10, eTVOC, NO, NO2, and CO2. Two data representations are provided. First, raw multivariate signal files contain 16 synchronized sensing axes per sample: NO2, O3, CO, NO, eTVOC, mPM1.0, mPM2.5, mPM4.0, mPM10, nPM0.5, nPM1.0, nPM2.5, nPM4.0, nPM10, tsPM, and CO2. Second, windowed L64 files provide fixed-length inference inputs formed from 64 samples across 16 axes, resulting in 1024 values per window.The dataset was constructed from nominal air-quality monitoring logs and poisoned attack variants generated for freeze and replay scenarios while preserving the remaining sensor channels as live multivariate context. It is intended for research on TinyML, embedded AI, cyber-physical security, attack classification, multivariate time-series analysis, and secure air-quality IoT systems."
Fields of Study & Subjects
Network Security and Intrusion Detection
IoT and Edge/Fog Computing
Adversarial Robustness in Machine Learning
Publication details
- Journal / source
- IEEE DataPort
- Publication date
- 27 March 2026
- Language
- EN
- Document type
- Dataset