Symbolic Execution for Attribution and Attack Synthesis in Neural Networks
| Title | Symbolic Execution for Attribution and Attack Synthesis in Neural Networks |
| Publication Type | Conference Paper |
| Year of Publication | 2019 |
| Authors | Gopinath, Divya, S. Pasareanu, Corina, Wang, Kaiyuan, Zhang, Mengshi, Khurshid, Sarfraz |
| Conference Name | 2019 IEEE/ACM 41st International Conference on Software Engineering: Companion Proceedings (ICSE-Companion) |
| Date Published | May 2019 |
| Publisher | IEEE |
| ISBN Number | 978-1-7281-1764-5 |
| Keywords | adversarial attacks, adversarial generation, attack synthesis, attribution, composability, core ideas, deep neural networks, DeepCheck lightweight symbolic analysis, DNN, DNN validation, Human Behavior, image classification, Image resolution, Importance Analysis, Metrics, neural nets, program analysis, pubcrawl, symbolic execution |
| Abstract | This paper introduces DeepCheck, a new approach for validating Deep Neural Networks (DNNs) based on core ideas from program analysis, specifically from symbolic execution. DeepCheck implements techniques for lightweight symbolic analysis of DNNs and applies them in the context of image classification to address two challenging problems: 1) identification of important pixels (for attribution and adversarial generation); and 2) creation of adversarial attacks. Experimental results using the MNIST data-set show that DeepCheck's lightweight symbolic analysis provides a valuable tool for DNN validation. |
| URL | https://ieeexplore.ieee.org/document/8802760/ |
| DOI | 10.1109/ICSE-Companion.2019.00115 |
| Citation Key | gopinath_symbolic_2019 |
- DNN validation
- Symbolic Execution
- pubcrawl
- program analysis
- neural nets
- Metrics
- Importance Analysis
- Image resolution
- image classification
- Human behavior
- adversarial attacks
- DNN
- DeepCheck lightweight symbolic analysis
- deep neural networks
- core ideas
- composability
- attribution
- attack synthesis
- adversarial generation