DenDrift: A Drift-Aware Algorithm for Host Profiling

Detecting and reacting to unauthorized actions is an essential task in security monitoring. What make this task challenging are the large number and various categories of hosts and processes to monitor. To these we should add the lack of an exact definition of normal behavior for each category. Host profiling using stream clustering algorithms is an effective means of analyzing hosts' behaviors, categorizing them, and identifying atypical ones. However, unforeseen changes in behavioral data (i.e. concept drift) make the obtained profiles unreliable. DenStream is a well-known stream clustering algorithm, which can be effectively used for host profiling. This algorithm is an incremental extension of DBSCAN which is a non-parametric algorithm widely used in real-world clustering applications. Recent experimental studies indicate that DenStream is not robust against concept drift. In this paper, we present DenDrift as a drift-aware host profiling algorithm based on DenStream. DenDrift relies on non-negative matrix factorization for dimensionality reduction and Page-Hinckley test for drift detection. We have done experiments on both synthetic and industrial datasets and the results affirm the robustness of DenDrift against abrupt, gradual and incremental drifts.

Multi-level Binarized LSTM in EEG Classification for Wearable Devices

Long Short-Term Memory (LSTM) is widely used in various sequential applications. Complex LSTMs could be hardly deployed on wearable and resourced-limited devices due to the huge amount of computations and memory requirements. Binary LSTMs are introduced to cope with this problem, however, they lead to significant accuracy loss in some application such as EEG classification which is essential to be deployed in wearable devices. In this paper, we propose an efficient multi-level binarized LSTM which has significantly reduced computations whereas ensuring an accuracy pretty close to full precision LSTM. By deploying 5-level binarized weights and inputs, our method reduces area and delay of MAC operation about 31* and 27* in 65nm technology, respectively with less than 0.01% accuracy loss. In contrast to many compute-intensive deep-learning approaches, the proposed algorithm is lightweight, and therefore, brings performance efficiency with accurate LSTM-based EEG classification to real-time wearable devices.

MuBiNN: Multi-Level Binarized Recurrent Neural Network for EEG signal Classification

Recurrent Neural Networks (RNN) are widely used for learning sequences in applications such as EEG classification. Complex RNNs could be hardly deployed on wearable devices due to their computation and memory-intensive processing patterns. Generally, reduction in precision leads much more efficiency and binarized RNNs are introduced as energy-efficient solutions. However, naive binarization methods lead to significant accuracy loss in EEG classification. In this paper, we propose a multi-level binarized LSTM, which significantly reduces computations whereas ensuring an accuracy pretty close to the full precision LSTM. Our method reduces the delay of the 3-bit LSTM cell operation 47* with less than 0.01% accuracy loss.