Abstract
Security is one of the major concerns in Industrial Wireless Sensor Networks (IWSNs). To assure the security in clustered IWSNs, this paper presents a secure clustering protocol with fuzzy trust evaluation and outlier detection (SCFTO). Firstly, to deal with the transmission uncertainty in an open wireless medium, an interval type-2 fuzzy logic controller is adopted to estimate the trusts. And then a density based outlier detection mechanism is introduced to acquire an adaptive trust threshold used to isolate the malicious nodes from being cluster heads. Finally, a fuzzy based cluster heads election method is proposed to achieve a balance between energy saving and security assurance, so that a normal sensor node with more residual energy or less confidence on other nodes has higher probability to be the cluster head. Extensive experiments verify that our secure clustering protocol can effectively defend the network against attacks from internal malicious or compromised nodes.
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores
Abstract
Machine learning models are typically deployed in a test setting that differs from the training setting, potentially leading to decreased model performance because of domain shift. If we could estimate the performance that a pre-trained model would achieve on data from a specific deployment setting, for example a certain clinic, we could judge whether the model could safely be deployed or if its performance degrades unacceptably on the specific data. Existing approaches estimate this based on the confidence of predictions made on unlabeled test data from the deployment's domain. We find existing methods struggle with data that present class imbalance, because the methods used to calibrate confidence do not account for bias induced by class imbalance, consequently failing to estimate class-wise accuracy. Here, we introduce class-wise calibration within the framework of performance estimation for imbalanced datasets. Specifically, we derive class-specific modifications of state-of-the-art confidence-based model evaluation methods including temperature scaling (TS), difference of confidences (DoC), and average thresholded confidence (ATC). We also extend the methods to estimate Dice similarity coefficient (DSC) in image segmentation. We conduct experiments on four tasks and find the proposed modifications consistently improve the estimation accuracy for imbalanced datasets. Our methods improve accuracy estimation by 18\% in classification under natural domain shifts, and double the estimation accuracy on segmentation tasks, when compared with prior methods.
Abstract
The explosion of global social media and online communication platforms has changed how we interact with each other and as a society, bringing with it new security and privacy challenges. Like all technologies, these platforms can be abused and they are routinely used to attempt to cause harm at scale. One of the most significant offence types that is enabled by these platforms is child sexual abuse - both scaling existing abuse and enabling entirely new types of online-only abuse where the impacts on the victim are equally catastrophic. Many platforms invest significantly in combating this crime, referring confirmed evidence of illegality to law enforcement. The introduction of end-to-end encryption and similar technologies breaks many of the mitigations in place today and this has led to a debate around the apparent dichotomy of good child safety and good general user privacy and security. This debate has concentrated on the problem of detecting offenders sharing known abuse imagery using a technique known as client side scanning. We will show that the real problem of online child sexual abuse is much more complex than offender image sharing, providing a new set of 'harm archetypes' to better group harms into categories that have similar technical characteristics and, as far as we are able, bring more clarity to the processes currently used by platforms and law enforcement in relation to child sexual abuse content and the real world impacts. We explore, at a high level, a variety of techniques that could be used as part of any potential solution and examine the benefits and disbenefits that may accrue in various use cases, and use a hypothetical service as an example of how various techniques could be brought together to provide both user privacy and security, while protecting child safety and enabling law enforcement action.
Evaluating the Stability of Deep Image Quality Assessment With Respect to Image Scaling
Authors: Koki Tsubota, Hiroaki Akutsu, Kiyoharu Aizawa
Subjects: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
Abstract
Image quality assessment (IQA) is a fundamental metric for image processing tasks (e.g., compression). With full-reference IQAs, traditional IQAs, such as PSNR and SSIM, have been used. Recently, IQAs based on deep neural networks (deep IQAs), such as LPIPS and DISTS, have also been used. It is known that image scaling is inconsistent among deep IQAs, as some perform down-scaling as pre-processing, whereas others instead use the original image size. In this paper, we show that the image scale is an influential factor that affects deep IQA performance. We comprehensively evaluate four deep IQAs on the same five datasets, and the experimental results show that image scale significantly influences IQA performance. We found that the most appropriate image scale is often neither the default nor the original size, and the choice differs depending on the methods and datasets used. We visualized the stability and found that PieAPP is the most stable among the four deep IQAs.
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores
Abstract
Machine learning models are typically deployed in a test setting that differs from the training setting, potentially leading to decreased model performance because of domain shift. If we could estimate the performance that a pre-trained model would achieve on data from a specific deployment setting, for example a certain clinic, we could judge whether the model could safely be deployed or if its performance degrades unacceptably on the specific data. Existing approaches estimate this based on the confidence of predictions made on unlabeled test data from the deployment's domain. We find existing methods struggle with data that present class imbalance, because the methods used to calibrate confidence do not account for bias induced by class imbalance, consequently failing to estimate class-wise accuracy. Here, we introduce class-wise calibration within the framework of performance estimation for imbalanced datasets. Specifically, we derive class-specific modifications of state-of-the-art confidence-based model evaluation methods including temperature scaling (TS), difference of confidences (DoC), and average thresholded confidence (ATC). We also extend the methods to estimate Dice similarity coefficient (DSC) in image segmentation. We conduct experiments on four tasks and find the proposed modifications consistently improve the estimation accuracy for imbalanced datasets. Our methods improve accuracy estimation by 18\% in classification under natural domain shifts, and double the estimation accuracy on segmentation tasks, when compared with prior methods.
Keyword: calibration
Differentiable Agent-based Epidemiology
Authors: Ayush Chopra, Alexander Rodríguez, Jayakumar Subramanian, Balaji Krishnamurthy, B. Aditya Prakash, Ramesh Raskar
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Populations and Evolution (q-bio.PE); Quantitative Methods (q-bio.QM)
Abstract
Mechanistic simulators are an indispensable tool for epidemiology to explore the behavior of complex, dynamic infections under varying conditions and navigate uncertain environments. ODE-based models are the dominant paradigm that enable fast simulations and are tractable to gradient-based optimization, but make simplifying assumptions about population homogeneity. Agent-based models (ABMs) are an increasingly popular alternative paradigm that can represent the heterogeneity of contact interactions with granular detail and agency of individual behavior. However, conventional ABM frameworks are not differentiable and present challenges in scalability; due to which it is non-trivial to connect them to auxiliary data sources easily. In this paper we introduce GradABM which is a new scalable, fast and differentiable design for ABMs. GradABM runs simulations in few seconds on commodity hardware and enables fast forward and differentiable inverse simulations. This makes it amenable to be merged with deep neural networks and seamlessly integrate heterogeneous data sources to help with calibration, forecasting and policy evaluation. We demonstrate the efficacy of GradABM via extensive experiments with real COVID-19 and influenza datasets. We are optimistic this work will bring ABM and AI communities closer together.
Cancer Subtyping by Improved Transcriptomic Features Using Vector Quantized Variational Autoencoder
Abstract
Defining and separating cancer subtypes is essential for facilitating personalized therapy modality and prognosis of patients. The definition of subtypes has been constantly recalibrated as a result of our deepened understanding. During this recalibration, researchers often rely on clustering of cancer data to provide an intuitive visual reference that could reveal the intrinsic characteristics of subtypes. The data being clustered are often omics data such as transcriptomics that have strong correlations to the underlying biological mechanism. However, while existing studies have shown promising results, they suffer from issues associated with omics data: sample scarcity and high dimensionality. As such, existing methods often impose unrealistic assumptions to extract useful features from the data while avoiding overfitting to spurious correlations. In this paper, we propose to leverage a recent strong generative model, Vector Quantized Variational AutoEncoder (VQ-VAE), to tackle the data issues and extract informative latent features that are crucial to the quality of subsequent clustering by retaining only information relevant to reconstructing the input. VQ-VAE does not impose strict assumptions and hence its latent features are better representations of the input, capable of yielding superior clustering performance with any mainstream clustering method. Extensive experiments and medical analysis on multiple datasets comprising 10 distinct cancers demonstrate the VQ-VAE clustering results can significantly and robustly improve prognosis over prevalent subtyping systems.
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores
Abstract
Machine learning models are typically deployed in a test setting that differs from the training setting, potentially leading to decreased model performance because of domain shift. If we could estimate the performance that a pre-trained model would achieve on data from a specific deployment setting, for example a certain clinic, we could judge whether the model could safely be deployed or if its performance degrades unacceptably on the specific data. Existing approaches estimate this based on the confidence of predictions made on unlabeled test data from the deployment's domain. We find existing methods struggle with data that present class imbalance, because the methods used to calibrate confidence do not account for bias induced by class imbalance, consequently failing to estimate class-wise accuracy. Here, we introduce class-wise calibration within the framework of performance estimation for imbalanced datasets. Specifically, we derive class-specific modifications of state-of-the-art confidence-based model evaluation methods including temperature scaling (TS), difference of confidences (DoC), and average thresholded confidence (ATC). We also extend the methods to estimate Dice similarity coefficient (DSC) in image segmentation. We conduct experiments on four tasks and find the proposed modifications consistently improve the estimation accuracy for imbalanced datasets. Our methods improve accuracy estimation by 18\% in classification under natural domain shifts, and double the estimation accuracy on segmentation tasks, when compared with prior methods.
Keyword: out of distribution detection
There is no result
Keyword: out-of-distribution detection
There is no result
Keyword: expected calibration error
There is no result
Keyword: overconfident
There is no result
Keyword: overconfidence
There is no result
Keyword: confidence
A Secure Clustering Protocol with Fuzzy Trust Evaluation and Outlier Detection for Industrial Wireless Sensor Networks
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores
Keyword: scaling
Thoughts on child safety on commodity platforms
Evaluating the Stability of Deep Image Quality Assessment With Respect to Image Scaling
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores
Keyword: calibration
Differentiable Agent-based Epidemiology
Cancer Subtyping by Improved Transcriptomic Features Using Vector Quantized Variational Autoencoder
Estimating Model Performance under Domain Shifts with Class-Specific Confidence Scores