No Longer Sleeping with a Bomb

Wang, Jingyuan; Chen, Chao; Wu, Junjie; Zhang, Xiong

doi:10.1145/3097983.3097985

Cited by 24 publications

(3 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, [9] predicted wildfire risk using dynamic temporal weather data. [10] leveraged heterogeneous big data to deal with the issue in dangerous goods transportation and [11] utilized dynamically updated data to identify the likelihood of a fire event.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Intelligent Urban Sensing for Gas Leakage Risk Assessment

et al. 2023

View full text Add to dashboard Cite

In order to mitigate gas leakage damage, it is important to assess systematic risk of gas leakage in utility networks. In practice, leakage detection systems heavily rely on tedious and expensive human efforts, such as manually checking and assessing the risk, and thus only a very limited portion of the gas pipeline network can usually be assessed during a period of time. Therefore, it is critical to develop tools for automatically and systematically evaluating the whole gas pipeline network and revealing high-risk sites for checking in a timely fashion. To this end, in this paper, we develop an intelligent gas leakage risk assessment system based on the analysis of large-scale multi-source data, such as gas pipeline data, Point of Interests (POIs) data, and human mobility data. However, it is a non-trivial endeavor to design such a risk assessment system due to the unclear leakage mechanism, complex environmental conditions, and large size of the pipeline network. To address these challenges, both internal features of gas pipelines and external environmental features should be exploited. Specifically, we design a novel urban sensing technique to extract environmental features by analyzing human mobility data. Then, a joint learning neural network is developed to assess the leakage risk by integrating both internal and external features. Moreover, an intelligent risk assessment system is implemented and deployed for experiments in real-world scenarios. The results demonstrate the effectiveness of our system for citywide gas leakage risk assessment. Indeed, the proposed system could help to locate the potential leakage sites in a timely way.INDEX TERMS Human mobility, urban sensing, risk management, gas leakage assessment I. INTRODUCTION

show abstract

Section: Related Workmentioning

confidence: 99%

“…A probabilistic model was trained to infer the population mobility in cities impacted by disasters [14]. [10] identified risky zones by matching the transportation trajectory of dangerous goods with human activity data. In the field of city management, [15] provided a systematic POI demand analysis with the help of taxi flow data.…”

Section: Related Workmentioning

confidence: 99%

Intelligent Urban Sensing for Gas Leakage Risk Assessment

et al. 2023

View full text Add to dashboard Cite

show abstract

“…As early proposed by Szegedy et al [28], it is very easy to mislead the outputs of a DL model by slightly perturbing input examples to form adversarial examples. This weakness causes DL models very unreliable in applications that are vulnerable to deliberate attacks, such as dangerous goods transportation [29] and military systems, and eventually could lead to substantial economic losses and even the costs of lives.…”

Section: Introductionmentioning

confidence: 99%

Interpretability is a Kind of Safety: An Interpreter-based Ensemble for Adversary Defense

Wang

Yan-ru

et al. 2020

Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery &Amp; Data Mining

Self Cite

View full text Add to dashboard Cite

While having achieved great success in rich real-life applications, deep neural network (DNN) models have long been criticized for their vulnerability to adversarial attacks. Tremendous research efforts have been dedicated to mitigating the threats of adversarial attacks, but the essential trait of adversarial examples is not yet clear, and most existing methods are yet vulnerable to hybrid attacks and suffer from counterattacks. In light of this, in this paper, we first reveal a gradient-based correlation between sensitivity analysisbased DNN interpreters and the generation process of adversarial examples, which indicates the Achilles's heel of adversarial attacks and sheds light on linking together the two long-standing challenges of DNN: fragility and unexplainability. We then propose an interpreter-based ensemble framework called X-Ensemble for robust adversary defense. X-Ensemble adopts a novel detectionrectification process and features in building multiple sub-detectors and a rectifier upon various types of interpretation information toward target classifiers. Moreover, X-Ensemble employs the Random Forests (RF) model to combine sub-detectors into an ensemble detector for adversarial hybrid attacks defense. The non-differentiable property of RF further makes it a precious choice against the counterattack of adversaries. Extensive experiments under various types of state-of-the-art attacks and diverse attack scenarios demonstrate the advantages of X-Ensemble to competitive baseline methods. CCS CONCEPTS• Computing methodologies → Neural networks.

show abstract