Abstract:We conduct a large‐scale empirical study of 11 predictive uncertainty metrics for DNNs on their effectiveness in detecting valid and invalid test inputs and find three best‐performance metrics: PE‐DP, EN and EN‐DP. We propose an uncertainty‐guided deep learning test approach to improve the naturalness of the generated test inputs. Extensive experiments show that our approach can generate more valid test inputs than the baselines, and appropriate data augmentation techniques can further boost performance. Despite numerous applications of deep learning technologies on critical tasks in various domains, advanced deep neural networks (DNNs) face persistent safety and security challenges, such as the overconfidence in predicting out‐of‐distribution samples and susceptibility to adversarial examples. Thorough testing by exploring the input space serves as a key strategy to ensure their robustness and trustworthiness of these networks. However, existing testing methods focus on disclosing more erroneous model behaviours, overlooking the validity of the generated test inputs. To mitigate this issue, we investigate devising valid test input generation method for DNNs from a predictive uncertainty perspective. Through a large‐scale empirical study across 11 predictive uncertainty metrics for DNNs, we explore the correlation between validity and uncertainty of test inputs. Our findings reveal that the predictive entropy‐based and ensemble‐based uncertainty metrics effectively characterize the input validity demonstration. Building on these insights, we introduce UCTest, an uncertainty‐guided deep learning testing approach, to efficiently generate valid and authentic test inputs. We formulate a joint optimization objective: to uncover the model's misbehaviours by maximizing the loss function and concurrently generate valid test input by minimizing uncertainty. Extensive experiments demonstrate that our approach outperforms the current testing methods in generating valid test inputs. Furthermore, incorporating natural variation through data augmentation techniques into UCTest effectively boosts the diversity of generated test inputs.

Uncertainty-Aware Prediction Validator in Deep Learning Models for Cyber-Physical System Data

Uncertainty-aware deep learning for reliable health monitoring in safety-critical energy systems

Prediction Surface Uncertainty Quantification in Object Detection Models for Autonomous Driving

Towards Safe Machine Learning for CPS: Infer Uncertainty from Training Data

Quantifying Uncertainty in Deep Learning Classification with Noise in Discrete Inputs for Risk-Based Decision Making

Uncertainty-aware deep learning for digital twin-driven monitoring: Application to fault detection in power lines

Predicting Safety Misbehaviours in Autonomous Driving Systems using Uncertainty Quantification

Fail-Safe Execution of Deep Learning based Systems through Uncertainty Monitoring

Uncertainty Quantification and Interval Prediction of Equipment Remaining Useful Life Based on Semisupervised Learning

An uncertainty-informed framework for trustworthy fault diagnosis in safety-critical applications

A Benchmark on Uncertainty Quantification for Deep Learning Prognostics

Assurance Monitoring of Cyber-Physical Systems with Machine Learning Components

Robustness Testing of Data and Knowledge Driven Anomaly Detection in Cyber-Physical Systems

Confidence Intervals and Simultaneous Confidence Bands Based on Deep Learning

Uncertainty-Aware Credit Card Fraud Detection Using Deep Learning

Validity Matters: Uncertainty‐Guided Testing of Deep Neural Networks

Probabilistic Deep Learning to Quantify Uncertainty in Air Quality Forecasting

RDeepSense: Reliable Deep Mobile Computing Models with Uncertainty Estimations

Distance Preserving Machine Learning for Uncertainty Aware Accelerator Capacitance Predictions

Reliable Deep Grade Prediction with Uncertainty Estimation

Uncertainty Modelling in Performability Prediction for Safety-Critical Systems