Attention-based Host Intrusion Detection System
Hierarchical attention model was used to detect intrusion from system call traces.
Project Duration: 2022-23
Highlights
- We proposed a hierarchical attention model for detecting intrusion from system call traces
- We stacked two bidirectional GRU layers to extract higher level features and achieved an AUC of 96%
Github Repository
Download Full Text
Abstract
With the development of deep learning, various method have been adopted in Host Intrusion Detection System or HIDS. However, the traditional methods of HIDS have been proven to be vulnerable to higher number of false alarm. In this study, we have proposed a novel hierarchical attention based deep learning method of detection intrusion on a host. We have evaluated our model on ADLF-LD, which is a collection of a trace data of Linux system calls. We have tuned our model’s hyper parameters to produce the optimum result, and our method has successfully outperforms the existing methods