Styx: A Data-Oriented Mutation Framework to Improve the Robustness of DNN
The robustness of deep neural network (DNN) is critical and challenging to ensure. In this paper, we propose a general data-oriented mutation framework, called Styx, to improve the robustness of DNN. Styx generates new training data by slightly mutating the training data. In this way, Styx ensures the DNN’s accuracy on the test dataset while improving the adaptability to small perturbations, i.e., improving the robustness. We have instantiated Styx for image classification and proposed pixel-level mutation rules that are applicable to any image classification DNNs. We have applied Styx on several commonly used benchmarks and compared Styx with the representative adversarial training methods. The preliminary experimental results indicate the effectiveness of Styx.
Tue 22 Sep Times are displayed in time zone: (UTC) Coordinated Universal Time
10:20 - 11:20: LBR + DS Poster (1)Late Breaking Results / Doctoral Symposium at Koala Chair(s): Kevin LeeDeakin University | |||
10:20 - 10:25 Poster | Efficient Multiplex Symbolic Execution with Adaptive Search Strategy Late Breaking Results Tianqi ZhangNational University of Defense Technology, Yufeng ZhangCollege of Information Science and Engineering, Hunan University, Zhenbang ChenCollege of Computer, National University of Defense Technology, Changsha, PR China, Ziqi ShuaiNational University of Defense Technology, Ji WangNational University of Defense Technology | ||
10:25 - 10:30 Poster | Styx: A Data-Oriented Mutation Framework to Improve the Robustness of DNN Late Breaking Results Meixi LiuNational University of Defense Technology, Changsha, China, Weijiang HongNational University of Defense Technology, Changsha, China, Weiyu PanNational University of Defense Technology, Changsha, China, Chendong FengCollege of Computer, National University of Defense Technology, Changsha, China, Zhenbang ChenCollege of Computer, National University of Defense Technology, Changsha, PR China, Ji WangNational University of Defense Technology | ||
10:30 - 10:35 Poster | Synthesizing Smart Solving Strategy for Symbolic Execution Late Breaking Results Zehua ChenNational University of Defense Technology, Zhenbang ChenCollege of Computer, National University of Defense Technology, Changsha, PR China, Ziqi ShuaiNational University of Defense Technology, Yufeng ZhangCollege of Information Science and Engineering, Hunan University, Weiyu PanNational University of Defense Technology, Changsha, China | ||
10:35 - 10:40 Poster | Privacy Assessment of Android Clipboard Late Breaking Results Zach Wei WangThe University of Adelaide, Ruoxi SunThe University of Adelaide, Jason Minhui XueThe University of Adelaide, Damith C. RanasingheThe University of Adelaide DOI | ||
10:40 - 10:45 Poster | The Symptom, Cause and Repair of Workaround Late Breaking Results Daohan SongShanghai Jiao Tong University, Hao ZhongShanghai Jiao Tong University, Li JiaShanghai Jiao Tong University | ||
10:45 - 10:50 Poster | Edge4Sys: A Device-Edge Collaborative Framework for MEC based Smart Systems Late Breaking Results Han GaoSchool of Computer Science and Technology, Anhui University, Yi XuSchool of Computer Science and Technology, Anhui University, Xiao LiuSchool of Information Technology, Deakin University, Jia XuSchool of Computer Science and Technology, Anhui University, Tianxiang ChenSchool of Computer Science and Technology, Anhui University, Bowen ZhouSchool of Computer Science and Technology, Anhui University, Rui LiSchool of Information Technology, Deakin University, Xuejun LiSchool of Computer Science and Technology, Anhui University | ||
10:50 - 10:55 Poster | Towards Immersive Comprehension of Software Systems Using Augmented Reality - An Empirical Evaluation Late Breaking Results Rohit MehraAccenture Labs, India, Vibhu Saujanya SharmaAccenture Labs, Bangalore, India, Vikrant KaulgudAccenture Labs, India, Sanjay PodderAccenture, Adam P. BurdenAccenture | ||
10:55 - 11:00 Poster | Towards Programming and Verification for Activity-Oriented Smart Home Systems Late Breaking Results Xuansong LiSchool of Computer Science and Engineering, Nanjing University of Science and Technology, Wei SongSchool of Computer Science and Engineering, Nanjing University of Science and Technology, Xiangyu ZhangPurdue University, USA | ||
11:00 - 11:05 Talk | Towards Robust Production Machine Learning Systems: Managing Dataset Shift Doctoral Symposium Hala AbdelkaderApplied Artificial Intelligence Institute, Deakin University | ||
11:05 - 11:10 Talk | Using Defect Prediction to Improve the Bug Detection Capability of Search-Based Software Testing Doctoral Symposium Anjana PereraMonash University DOI Pre-print |