ACM SIGSOFT Distinguished Paper Award

Game testing has been long recognized as a notoriously challenging task, which mainly relies on manual playing and scripting based testing in game industry. Even until recently, automated game testing still remains to be largely untouched niche. A key challenge is that game testing often requires to play the game as a sequential decision process. A bug may only be triggered until completing certain difficult intermediate tasks, which requires a certain level of intelligence. The recent success of deep reinforcement learning (DRL) sheds light on advancing automated game testing, without human competitive intelligent support. However, the existing DRLs mostly focus on winning the game rather than game testing. To bridge the gap, in this paper, we first perform an in-depth analysis of 1349 real bugs from four real-world commercial game products. Based on this, we propose four oracles to support automated game testing, and further propose Wuji, an on-the-fly game testing framework, which leverages evolutionary algorithm, DRL and multi-objective optimization to perform automatic game testing. Wuji balances between winning the game and exploring the space of the game. Winning the game allows the agent to make progress in the game, while space exploration increases the possibility of discovering bugs. We conduct a large-scale evaluation on a simple game and two popular commercial games. The results demonstrate the effectiveness of Wuji in exploring space and detecting bugs. Moreover, Wuji found 3 previously unknown bugs, which have been confirmed by the developers, in the commercial games.

Yan Zheng

Tianjin University

Singapore

Xiaofei Xie

Nanyang Technological University

Ting Su

ETH Zurich

Lei Ma

Kyushu University

Japan

Jianye Hao

Tianjin University

Zhaopeng Meng

Tianjin University

Yang Liu

Nanyang Technological University, Singapore

Singapore

Ruimin Shen

Fuxi AI Lab in Netease

Yinfeng Chen

Fuxi AI Lab in Netease

Changjie Fan

Fuxi AI Lab in Netease

Due to the popularity of deep learning (DL) applications, testing DL libraries are becoming more and more important. Different from traditional testing, for which output is asserted definitely (e.g., an output is compared with an oracle for equality), testing deep learning libraries often requires to perform oracle approximations, i.e., the output is allowed to be within a restricted range of the oracle. However, oracle approximations have not been studied in prior empirical work that focuses on traditional testing practices. The prevalence, common practices, evolution, and maintenance challenges of oracle approximations remain unknown. In this work, we studied oracle approximation assertions that are implemented in four popular deep learning libraries. Our study shows that oracle approximation assertions are a significant portion among all the assertions in the test suites of deep learning libraries. We identify the commonly-used oracle types when there are approximations being performed on oracles through a comprehensive manual study. In addition, we find that developers frequently modify code on oracle approximations, i.e., using a different approximation API, modifying the oracle or the output from the code under test, and using a different threshold value. Finally, we performed in-depth studies to understand the reasons behind the evolution of oracle approximation assertions and our findings reveal maintenance challenges.

Mahdi Nejadgholi

Concordia University

Jinqiu Yang

Concordia University, Montreal, Canada

We present techniques for automatically inferring formal properties of feed-forward neural networks. We observe that a significant part (if not all) of the logic of feed forward networks is captured in the activation status (on or off) of its neurons. We propose to extract patterns based on neuron decisions as preconditions that imply certain desirable output property e.g., the prediction being a certain class. Together, the inferred preconditions and the output property form a {\em contract} for the network. We present techniques to extract input properties} encoding convex predicates on the input space that imply given output properties and layer properties, representing network properties captured in the hidden layers that imply the desired output behavior. We apply our techniques on networks for the MNIST and ACASXU applications. Our experiments highlight the use of the inferred properties in a variety of tasks, such as explaining predictions, providing robustness guarantees, simplifying proofs, and network distillation.

Divya Gopinath

Carnegie Mellon University

Hayes Converse

The University of Texas at Austin

Corina S. Pasareanu

Carnegie Mellon University Silicon Valley, NASA Ames Research Center

Ankur Taly

Google

Deep Learning (DL) has recently achieved tremendous success in various application domains. A variety of DL frameworks and platforms play a key role to catalyze such progress. However, the differences in architecture designs and implementations of existing frameworks and platforms bring new challenges for DL software development and deployment. Till now, there is no study on how various mainstream frameworks and platforms influence both DL software development and deployment in practice.

To fill this gap, we take the first step towards understanding how the most widely-used DL frameworks and platforms support the DL software development and deployment. We conduct a systematic study on these frameworks and platforms by using two types of DNN architectures and three popular datasets. (1) For development process, we investigate the prediction accuracy under the same runtime training configuration or same model weights/biases. We also study the adversarial robustness of trained models by leveraging the existing adversarial attack techniques.The experimental results show that the {computing differences} across frameworks could result in an obvious prediction accuracy decline, which should draw the attention of DL developers. (2) For deployment process, we investigate the prediction accuracy and performance (refers to time cost and memory consumption) when the trained models are migrated/quantized from PC to real mobile devices and web browsers. The DL platform study unveils that the migration and quantization still suffer from compatibility and reliability issues. Meanwhile, we find several DL software bugs by using the results as a benchmark. We further validate the results through bug confirmation from stakeholders and industrial positive feedback to highlight the implications of our study. Through our study, we summarize practical guidelines, identify challenges and pinpoint new research directions, such as understanding the characteristics of DL frameworks and platforms, avoiding compatibility/reliability issues, detecting DL software bugs, and reducing time cost and memory consumption towards developing and deploying high quality DL systems effectively.

Qianyu Guo

Tianjin University

Sen Chen

Nanyang Technological University, Singapore

Singapore

Xiaofei Xie

Nanyang Technological University

Lei Ma

Kyushu University

Japan

Qiang Hu

Kyushu University, Japan

Hongtao Liu

Tianjin University

Yang Liu

Nanyang Technological University, Singapore

Singapore

Jianjun Zhao

Kyushu University

Japan

Li Xiaohong

TianJin University

China

Deep neural networks~(DNNs) are increasingly expanding their real-world applications across domains, e.g., image processing, speech recognition and natural language processing. However, there is still limited tool support for DNN testing in terms of test data quality and model robustness. In this paper, we introduce a mutation testing-based tool for DNNs, DeepMutation++, which facilitates the DNN quality evaluation, supporting both feed-forward neural networks~(FNNs) and stateful recurrent neural networks~(RNNs). It not only enables to statically analyze the robustness of a DNN model against the input as a whole, but also allows to identify the vulnerable segments of a sequential input (e.g. audio input) by runtime analysis. It is worth noting that DeepMutation++ specially features the support of RNNs mutation testing. The tool demo video can be found on the project website https://sites.google.com/view/deepmutationpp.

Qiang Hu

Kyushu University, Japan

Lei Ma

Kyushu University

Japan

Xiaofei Xie

Nanyang Technological University

Bing Yu

Kyushu University, Japan

Yang Liu

Nanyang Technological University, Singapore

Singapore

Jianjun Zhao

Kyushu University

Japan

Deep neural network (DNN) has been widely applied to safety-critical scenarios such as autonomous vehicle, security surveillance, and cyber-physical control systems. Yet, the incorrect behaviors of DNNs can lead to severe accidents and tremendous losses due to hidden defects. In this paper, we present DeepHunter, a general-purpose fuzzing framework for detecting defects of DNNs. DeepHunter is inspired by traditional grey-box fuzzing and aims to increase the overall test coverage by applying adaptive heuristics according to runtime feedback. Specifically, DeepHunter provides a series of seed selection strategies, metamorphic mutation strategies, and testing criteria customized to DNN testing; all these components support multiple built-in configurations which are easy to extend. We evaluated DeepHunter on two popular datasets and the results demonstrate the effectiveness of DeepHunter in achieving coverage increase and detecting real defects. A video demonstration which showcases the main features of DeepHunter can be found at https://youtu.be/s5DfLErcgrc.

Xiaofei Xie

Nanyang Technological University

Hongxu Chen

Nanyang Technological University

Yi Li

Nanyang Technological University

Singapore

Lei Ma

Kyushu University

Japan

Yang Liu

Nanyang Technological University, Singapore

Singapore

Jianjun Zhao

Kyushu University

Japan