Speaker: Dr. Kun Yang
From: Princeton University
Abstract
Novelty detection is important in the Internet of Things (“IoT”) due to the potential threats that IoT devices can present. One-Class Support Vector Machines (OCSVMs) are one of the common approaches for novelty detection due to their ability to identify a wide range of nonlinear classification boundaries. Such flexibility is appropriate for IoT devices and applications, which exhibit complexity due to the vast heterogeneity of devices and the wide range of traffic patterns under different operating modalities.
However, conventional OCSVMs can introduce prohibitive memory and computational overhead in detection. This work designs, implements, and evaluates an efficient OCSVM for such practical settings. We extend Nyström and (Gaussian) Sketching approaches to OCSVM, combining these methods with clustering and Gaussian mixture models to achieve 15-30x speedup in prediction time and 30-40x reduction in memory requirements without sacrificing detection accuracy.
For more info, please follow this link.
Read More