This research paper evaluates the effectiveness of adversarial training (AT) as a defense mechanism against poisoning attacks in online AutoML pipelines for Internet of Things (IoT) networks. The study specifically examined label flip and noise injection attacks on various streaming-capable AutoML learners, including Hoeffding Tree, Leveraging Bagging, Adaptive Random Forest, Hoeffding Adaptive Tree, and Streaming Random Patches. Results indicated that AT-Streaming Random Patches achieved the highest F1-score (0.904) against label flip poisoning, while AT-Leveraging Bagging performed best (0.933) against noise injection poisoning. AI
IMPACT This research highlights potential vulnerabilities in IoT machine learning systems and evaluates defense mechanisms against data poisoning attacks.
RANK_REASON This is a research paper detailing an evaluation of defense strategies against machine learning attacks. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
- Adaptive Random Forest
- Hoeffding Adaptive Tree
- Hoeffding Tree Algorithms for Anomaly Detection in Streaming Datasets: A Survey
- Hugging Face
- Internet of Things
- Leveraging Bagging for Evolving Data Streams
- machine learning
- Streaming Random Patches
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
