Junyao Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Junyao Huang has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Chenhui Lu
Tsinghua University
1 shared publication - GuoLou Ping
Tsinghua University
1 shared publication Lin Sun
Tsinghua University
1 shared publication- Xiaojun Ye
School of Software, Tsinghua University, Beijing, China
1 shared publication
Latest Publications
- 6 May 2020
- Published by Springer Nature
- Vol. 12085, 178-190
TCN-ATT: A Non-recurrent Model for Sequence-Based Malware Detection
Malware detection based on API call sequences is widely used for the ability to model program behaviours. But RNN-based models for this task usually have bottlenecks in efficiency and accuracy due to their recurrent structure. In this paper, we propose a Temporal Convolutional Network with ATTention (TCN-ATT) architecture, which processes sequences with high parallelization and is robust to sequence length. The proposed TCN-ATT consists of three components: (1) a TCN module which processes sequence with convolutional structure, (2) an attention layer to select effective features and (3) a split-and-combine mechanism to fit inputs with various size. A formalized deduplication method is also proposed to reduce redundancy with less information loss. According to our experiments, the proposed model reaches an accuracy of 98.60% and reduces time cost by over 60% compared with existing RNN-based models.
- 6 May 2020
- Published by Springer Nature
- Vol. 12085, 178-190
TCN-ATT: A Non-recurrent Model for Sequence-Based Malware Detection
Malware detection based on API call sequences is widely used for the ability to model program behaviours. But RNN-based models for this task usually have bottlenecks in efficiency and accuracy due to their recurrent structure. In this paper, we propose a Temporal Convolutional Network with ATTention (TCN-ATT) architecture, which processes sequences with high parallelization and is robust to sequence length. The proposed TCN-ATT consists of three components: (1) a TCN module which processes sequence with convolutional structure, (2) an attention layer to select effective features and (3) a split-and-combine mechanism to fit inputs with various size. A formalized deduplication method is also proposed to reduce redundancy with less information loss. According to our experiments, the proposed model reaches an accuracy of 98.60% and reduces time cost by over 60% compared with existing RNN-based models.