DSML 2021
Dependable and Secure Machine Learning


Workshop Program - Monday, 21 June 2021

14:30 CET Welcome to DSN-DSML 2021
Guanpeng(Justin) Li, University of Iowa, Hui Xu, Fudan University
Session 1: Keynote Talk
Session Chair: Homa Alemzadeh
14:35 CET

15:05 CET
Intelligent Software Engineering: Working at the Intersection of AI and Software Engineering
Tao Xie, Peking University
Q&A
15:10 CET Break
Session 2: Dependability Analysis of ML Models
Session Chair: Pinjia He
15:20 CET



15:30 CET



15:40 CET
Fault-Tolerant Low-Precision DNNs using Explainable AI
Muhammad Sabih, Frank Hannig, Jürgen Teich


Detecting Deep Neural Network Defects with Data Flow Analysis
Jiazhen Gu, Huanlin Xu, Yangfan Zhou, Xin Wang


Poisoning Attacks via Generative Adversarial Text to Image Synthesis
Keshav Kasichainula, Hadi Mansourifar, Weidong Shi


15:50 CET Break
Session 3: Keynote Talk
Session Chair: Karthik Pattabiraman
16:00 CET

16:30 CET
Achieving Error Resiliency for Machine Learning Models
Michael Paulitsch, Intel Labs
Q&A
16:35 CET Break
Session 4: Fault-Tolerant ML Systems
Session Chair: Varun Chandrasekaran
16:45 CET



16:55 CET



17:05 CET



17:15 CET
RADICS: Runtime Assurance of Distributed Intelligent Control Systems
Brian Wheatman, Jerry Chen, Tamim Sookoor, Yair Amir


An Approach for Peer-to-Peer Federated Learning
Tobias Wink, Zoltan Nochta


Byzantine Fault-Tolerant Distributed Machine Learning using D-SGD and Norm-Based Comparative Gradient Elimination (CGE)
Nirupam Gupta, Shuo Liu, Nitin Vaidya


A Queueing Analysis of Multi-model Multi-input Machine Learning Systems
Yuta Makino, Tuan Phung-Duc, Fumio Machida


17:25 CET Discussion and Closing

Keynotes


Intelligent Software Engineering: Working at the Intersection of AI and Software Engineering
Tao Xie, Peking University

Abstract: As an example of exploiting the synergy between AI and software engineering, the field of intelligent software engineering has emerged with various advances in recent years. Such a field broadly addresses issues on intelligent [software engineering] and [intelligence software] engineering. The former, intelligent [software engineering], focuses on instilling intelligence in approaches developed to address various software engineering tasks to accomplish high effectiveness and efficiency. The latter, [intelligence software] engineering, focuses on addressing various software engineering tasks for intelligence software, e.g., AI software. This talk will discuss recent research and future directions in the field of intelligent software engineering.

Speaker Bio: Tao Xie is a Chair Professor in the Department of Computer Science and Technology at Peking University, Beijing, China. He received an NSF CAREER Award, ACM SIGSOFT Distinguished Service Award, IEEE CS TCSE Distinguished Service Award, and various industrial faculty awards and distinguished/best paper awards. He is a co-Editor-in-Chief of the Wiley journal of Software Testing, Verification and Reliability (STVR). He served as the ISSTA 2015 Program Chair, Tapia 2017/2018 Program/General Chair, and an ICSE 2021 Program Co-Chair. He was selected by Lero as a David Lorge Parnas Fellow in 2019. He was selected as an ACM Distinguished Scientist in 2015, an IEEE Fellow in 2018, and an AAAS Fellow in 2019.


Achieving Error Resiliency for Machine Learning Models
Michael Paulitsch, Intel

Abstract: Modern computers and neural network accelerators play a major role in modern chip design. The talks looks at different failures mode of platform faults and presents potential failure impact of these faults. It presents fault and motivation on how to mitigate those faults. It also presents ways to monitor and mitigate these faults at different levels. With such approaches machine learning systems can achieve more robust performance and monitoring can achieve additional integrity.

Speaker Bio: Michael brings 20 years of work theoretical and applied research and technology work at university and different industries (aerospace, railway, automotive) in dependability in safety-critical and real-time systems including security aspects of all types. Michael fills the role of a Dependability Systems Architect (Principal Engineer) at Intel, Munich, Germany, as part Intel Labs Europe, since 2018. He pursues Dependable Artificial Intelligence and Machine Learning systems (resiliency) evaluates and ensures safe and dependable use of neural network models in safety-critical systems. He also looks at novel safety monitoring approaches at different system levels (chip, platform, application). From 2014 to 2018, he has been senior engineering manager and product line manager for Vital Platform (safety-critical computing and communication platform with security requirements) at Thales Ground Transportation Systems in Vienna, Austria. In these roles he has been responsible for the execution and strategy of as well as research for this vital platform. Before this, Michael has been Senior Expert of “Dependable Computing and Networks” as well as Scientific Director at Airbus corporate research in Munich, Germany. From 2003 to 2008, he worked at Honeywell Aerospace in the U.S. on software and electronic platforms in the area of business, regional, air transport, and human space avionics and engine control electronics. Michael has also been assistant professor at Technische Universitaet Wien, Vienna, Austria, 1997 to 2003.