Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper evaluates TabPFNv2.5, a transformer-based tabular foundation model, for IoT intrusion detection against traditional ensemble methods like Random Forest. Results on the TON IoT dataset show that TabPFNv2.5 achieves 40x faster inference speed compared to Random Forest, while maintaining 97% binary classification accuracy. The authors propose a hybrid pipeline leveraging TabPFNv2.5 for rapid threat screening and ensemble models for detailed classification, highlighting the challenges of detecting scanning attacks and the importance of feature similarity for cross-device generalization.
Tabular foundation models can slash inference time by 40x in IoT intrusion detection without sacrificing accuracy, opening the door for real-time smart city security operations.
Security operations in smart cities demand detection systems that balance accuracy with response time. While ensemble methods like Random Forest achieve high accuracy, their computational overhead impedes real-time forensic triage. We present the first systematic evaluation of TabPFNv2.5, a transformer-based foundation model, against traditional ensemble classifiers for IoT intrusion detection. Using the TON IoT dataset, we demonstrate that TabPFNv2.5 achieves 40 faster inference than Random Forest while maintaining 97% binary classification accuracy. We propose a hybrid pipeline in which TabPFNv2.5 performs rapid threat screening, while ensemble models handle detailed classification. Our analysis reveals that scanning attacks remain the hardest to detect (F1: 69.8%) and cross-device generalization depends critically on feature similarity. These findings establish foundation models as viable components for time-sensitive IoT security operations