检索结果-内蒙古大学图书馆

Analyze and Improve Differentially Private Federated Learning: A model robustness Perspective

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY 2025年 20卷 807-821页

作者： Zhang, Shuaishuai Huang, Jie Li, Peihao Southeast Univ Sch Cyber Sci & Engn Nanjing 211189 Peoples R China Purple Mt Labs Nanjing 211111 Peoples R China

Differentially Private Federated learning (DPFL) applies differential privacy (DP) techniques to preserve clients' privacy in Federated Learning (FL). Existing methods based on Gaussian Mechanism require the operations of model updates clipping and noise injection, which lead to a serious degradation in model accuracies. Several improved methods are proposed to mitigate the accuracy degradation by decreasing the scale of the injected noise. Different from previous methods, we firstly propose to enhance the model robustness against the DP noise for the accuracy improvement. In this paper, we develop a novel FL scheme with improved model robustness, called FedIMR, which can provide the client-level DP guarantee while maintaining a high model accuracy. We find that the injected noise leads to the fluctuation of loss values in the local training, hindering the model convergence seriously. This motivates us to improve the model robustness for narrowing down the bias of model outputs caused by the noise. The model robustness is evaluated with the signal-to-noise ratio (SNR) of each layer's outputs. Two techniques are proposed to improve the output SNR, including the logit vector normalization (LVN) and dynamic clipping threshold (DCT). Specifically, LVN normalizes the logit vertor to make the optimization algorithm keep increasing the model output, which is the signal item of the output SNR. DCT dynamically adjusts the clipping threshold to reduce the noise item of the output SNR. We also provide the privacy analysis and convergence results. Experiments are conducted over three famous datasets to evaluate the effectiveness of our method. Both the theoretical results and empirical experiments confirm that our FedIMR can achieve a better accuracy-privacy tradeoff than previous methods.

关键词： Noise robustness Privacy Training Convergence Computational modeling Signal to noise ratio Accuracy Discrete cosine transforms Differential privacy Federated learning differential privacy privacy preserving model robustness

来源：评论

学校读者我要写书评

暂无评论

Adversarial Vulnerabilities in Ransomware Detection: Enhancing Machine Learning model robustness 22

Adversarial Vulnerabilities in Ransomware Detection: Enhanci...

引用

22nd IEEE Consumer Communications and Networking Conference, CCNC 2025

作者： Heydari, Vahid Nyarko, Kofi Morgan State University Computer Science Department Baltimore United States Morgan State University Electrical and Computer Engineering Department Baltimore United States

ISBN: (纸本)9798331508050

Machine learning (ML) models are increasingly employed in cybersecurity to detect and prevent malicious activities such as ransomware attacks. However, these models are vulnerable to adversarial attacks, where subtle modifications to input data can cause a model to make incorrect predictions, potentially allowing threats to evade detection. This paper investigates the robustness of an ML model used for ransomware detection, demonstrating how adversarially modified samples can lead to a complete failure in detection despite high initial accuracy. We explore the effectiveness of retraining the model with these adversarial samples to restore its detection capabilities. Our findings reveal that while retraining can improve the model's performance, it also introduces new dependencies that may be exploited in future attacks. The results underscore the need for continuous evaluation and adaptation of ML models in cybersecurity, particularly in the face of evolving threats. This study highlights the importance of developing more resilient models and offers insights into potential strategies for enhancing the robustness of ML systems in cybersecurity applications. © 2025 IEEE.

关键词： Adversarial Attacks Cybersecurity Feature Importance Machine Learning model robustness Ransomware Detection

来源：评论

学校读者我要写书评

暂无评论

model robustness improvement by absorption and reduced scattering spectra in short wave near infrared spectral region

引用

BIOSYSTEMS ENGINEERING 2018年 176卷 114-124页

作者： He, Xueming Fu, Xiaping Rao, Xiuqin Zhejiang Univ Coll Biosyst Engn & Food Sci 866 Yuhangtang Rd Hangzhou 310058 Zhejiang Peoples R China Zhejiang Sci Tech Univ Fac Mech Engn & Automat 928 Second Ave Hangzhou 310018 Zhejiang Peoples R China Minist Agr Key Lab Site Proc Equipment Agr Prod Beijing Peoples R China

model robustness has always been the research focus for near infrared spectroscopy technique. In this study, we compared the model performance of four different kinds of spectra (transmittance, reflectance, absorption (mu(a)) and reduced scattering (mu'(s)))for the prediction of fat and protein of milk. The protein content of milk in prediction set was totally out of range of fat calibration set. The same happens in the case of the fat content of milk in prediction set and protein calibration set. The aim of this present study was to investigate the effect of the difference of protein content on the robustness of four fat calibration models, and of the difference of fat content on the robustness of four protein calibration models. The single integrating sphere system was first calibrated by using 90 liquid phantoms. The detection accuracy of the calibrated system was externally validated by another 80 phantoms. It turned out that the system can measure the absorption and reduced scattering coefficients with relatively high precision in the range of 780-1073 nm, which gave mean relative errors less than 12% and 5% for mu(a) and mu'(s) respectively. Partial least squares regression (PLSR) and stepwise multiple linear regression (SMLR) models were established, and the results showed that calibration model based on reduced scattering spectra could predict fat with the highest robustness for both regression methods, while the absorption spectra could predict protein content with the highest robustness by using SMLR. The separation of absorption and reduced scattering has great potential in improving the robustness of calibration models. (C) 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.

关键词： Milk Fat Protein Absorption Reduced scattering model robustness

来源：评论

学校读者我要写书评

暂无评论

model robustness in economics: the admissibility and evaluation of tractability assumptions

引用

SYNTHESE 2022年第1期200卷 32页

作者： O'Loughlin, Ryan Li, Dan Indiana Univ Dept Hist & Philosophy Sci & Med Bloomington IN 47405 USA Indiana Univ Dept Informat Bloomington IN 47405 USA

Lisciandra (2017) poses a challenge for robustness analysis (RA) as applied to economic models. She argues that substituting tractability assumptions risks altering the main mathematical structure of the model, thereby preventing the possibility of meaningfully evaluating the same model under different assumptions. In such cases RA is argued to be inapplicable. However, Lisciandra is mistaken to take the goal of RA as keeping the mathematical properties of tractability assumptions intact. Instead, RA really aims to keep the modeling component while varying the corresponding mathematical formulation. Thus, her argument concerning whether the associated mathematical properties of certain assumptions can be kept intact is irrelevant to the success of RA. Furthermore, we explicate and develop Lloyd's account of "model robustness" (Lloyd 2015) to provide solutions to Lisciandra's challenges. Our solutions are, namely, error analysis and independent empirical support. We conclude that although complex economic models do face potential dangers, there are solutions and robustness analysis need not be given up. (157).

关键词： robustness analysis model robustness Tractability assumptions model holism Economic models

来源：评论

学校读者我要写书评

暂无评论

model robustness of finite state nonlinear filtering over the infinite time horizon

引用

ANNALS OF APPLIED PROBABILITY 2007年第2期17卷 688-715页

作者： Chigansky, Pavel van Handel, Ramon Weizmann Inst Sci Dept Math IL-76100 Rehovot Israel CALTECH Phys Measurement & Control 26633 Pasadena CA 91125 USA

We investigate the robustness of nonlinear filtering for continuous time finite state Markov chains, observed in white noise, with respect to misspecification of the model parameters. It is shown that the distance between the optimal filter and that with incorrect model parameters converges to zero uniformly over the infinite time interval as the misspecified model converges to the true model, provided the signal obeys a mixing condition. The filtering error is controlled through the exponential decay of the derivative of the nonlinear filter with respect to its initial condition. We allow simultaneously for misspecification of the initial condition, of the transition rates of the signal, and of the observation function. The first two cases are treated by relatively elementary means, while the latter case requires the use of Skorokhod integrals and tools of anticipative stochastic calculus.

关键词： nonlinear filtering filter stability model robustness error bounds Markov chains anticipative stochastic calculus

来源：评论

学校读者我要写书评

暂无评论

Re-thinking model robustness from stability: a new insight to defend adversarial examples

引用

MACHINE LEARNING 2022年第7期111卷 2489-2513页

作者： Zhang, Shufei Huang, Kaizhu Xu, Zenglin Univ Liverpool Dept Comp Sci Liverpool L69 3BX Merseyside England Xian Jiaotong Liverpool Univ Dept Intelligent Sci Suzhou 215123 Peoples R China Shanghai Artificial Intelligence Lab 37th FloorAI Tower701 Yunjin Rd Shanghai Peoples R China Duke Kunshan Univ Data Sci Res Ctr Duke Ave 8 Suzhou 215316 Peoples R China Harbin Inst Technol Dept Comp Sci & Technol Shenzhen 518055 Guangdong Peoples R China

We study the model robustness against adversarial examples, referred to as small perturbed input data that may however fool many state-of-the-art deep learning models. Unlike previous research, we establish a novel theory addressing the robustness issue from the perspective of stability of the loss function in the small neighborhood of natural examples. We propose to exploit an energy function to describe the total variation in a small neighborhood and prove that reducing such energy guarantees the robustness against adversarial examples. We also show that the traditional training methods including adversarial training and virtual adversarial training tend to minimize the lower bound of our proposed energy function. Importantly, we prove that minimizing the energy function can obtain a better generalization bound than traditional adversarial training approaches. Through a series of experiments, we demonstrate the superiority of our model on different datasets for defending adversarial attacks. In particular, our proposed adversarial framework achieves the best performance compared with previous adversarial training methods on benchmark datasets CIFAR-10, CIFAR-100 and SVHN and they demonstrate much better robustness against adversarial examples than all the other comparison methods.

关键词： model robustness Adversarial examples Adversarial training Energy Adversarial generalization

来源：评论

学校读者我要写书评

暂无评论

Calibration of empirical models considering model fidelity and model robustness - Focusing on predictions of liquefaction-induced settlements

引用

ENGINEERING GEOLOGY 2016年 203卷 168-177页

作者： Gong, Wenping Tien, Yong Ming Hsein-Juang, C. Martin, James R., II Zhang, Jie Clemson Univ Glenn Dept Civil Engn Clemson SC 29634 USA Natl Cent Univ Dept Civil Engn Taoyuan 320 Taiwan Tongji Univ Dept Geotech Engn Shanghai 200092 Peoples R China

Most data-driven empirical models adopted in the geotechnical design have various degrees of uncertainty. Consequently, it is important to properly calibrate this uncertainty prior to its application in the geotechnical analysis to ensure the integrity of the final design. The conventional probabilistic model calibration approaches, such as maximum likelihood estimate and Bayesian method, focus only on model fidelity (measured through, for example, likelihood). These approaches often yield models with a large model uncertainty, especially when the observed data are scattered. A large model uncertainty leads to a large variation in the model prediction, which poses a significant challenge in a geotechnical design. In this paper, we propose a new model calibration approach that can consider both model fidelity and model robustness. The new model calibration approach offers a way to balance the objectives of model fidelity and model robustness. Because model fidelity and model robustness are two conflicting objectives, the new approach involves a multi-objective optimization that leads to a Pareto front, which defines a tradeoff relationship between model fidelity and model robustness. By enforcing a certain level of robustness, the variation in the model prediction can be reduced, which overcomes the major weakness of the traditional probabilistic approaches that focus solely on model fidelity. The new approach is demonstrated through applications to the problems of liquefaction-induced settlement prediction. (C) 2015 Elsevier B.V. All rights reserved.

关键词： Empirical models Liquefaction model fidelity model robustness Multi-objective optimization Probabilistic model calibration Settlement

来源：评论

学校读者我要写书评

暂无评论

Enhancing model robustness and Accuracy Against Adversarial Attacks via Adversarial Input Training

引用

INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS 2024年第3期15卷 1210-1228页

作者： Ingle, Ganesh Pawale, Sanjesh Vishwakarma Univ Dept Comp Engn Pune Maharashtra India

Adversarial attacks present a formidable challenge to the integrity of Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models, particularly in the domain of power quality disturbance (PQD) classification, necessitating the development of effective defense mechanisms. These attacks, characterized by their subtlety, can significantly degrade the performance of models critical for maintaining power system stability and efficiency. This study introduces the concept of adversarial attacks on CNN-LSTM models and emphasizes the critical need for robust *** propose Input Adversarial Training (IAT) as a novel defense strategy aimed at enhancing the resilience of CNN-LSTM models. IAT involves training models on a blend of clean and adversarially perturbed inputs, intending to improve their robustness. The effectiveness of IAT is assessed through a series of comparisons with established defense mechanisms, employing metrics such as accuracy, precision, recall, and F1-score on both unperturbed and adversarially modified *** results are compelling: models defended with IAT exhibit remarkable improvements in robustness against adversarial attacks. Specifically, IAT-enhanced models demonstrated an increase in accuracy on adversarially perturbed data to 85%, a precision improvement to 86%, a recall rise to 85%, and an F1-score enhancement to 85.5%. These figures significantly surpass those achieved by models utilizing standard adversarial training (75% accuracy) and defensive distillation (70% accuracy), showcasing IAT's superior capacity to maintain model accuracy under adversarial *** conclusion, IAT stands out as an effective defense mechanism, significantly bolstering the resilience of CNN-LSTM models against adversarial perturbations. This research not only sheds light on the vulnerabilities of these models to adversarial attacks but also establishes IAT as a benchmark in defense strategy development, promising enhanced security and re

关键词： Adversarial attacks Input Adversarial Training (IAT) deep learning security model robustness

来源：评论

学校读者我要写书评

暂无评论

Improving model robustness of traffic crash risk evaluation via adversarial mix-up under traffic flow fundamental diagram

引用

ACCIDENT ANALYSIS AND PREVENTION 2024年 194卷 107360页

作者： Yu, Rongjie Han, Lei Abdel-Aty, Mohamed Wang, Liqiang Zou, Zihang Minist Educ Key Lab Rd & Traff Engn 4800 Caoan Rd Shanghai 201804 Peoples R China Univ Cent Florida Dept Civil Environm & Construct Engn Orlando FL 32816 USA Univ Cent Florida Dept Comp Sci 32816HEC 437E Orlando FL 32816 USA

Recent state-of-art crash risk evaluation studies have exploited deep learning (DL) techniques to improve performance in identifying high-risk traffic operation statuses. However, it is doubtful if such DL-based models would remain robust to real-world traffic dynamics (e.g., random traffic fluctuations.) as DL models are sensitive to input changes, where small perturbations could lead to wrong predictions. This study raises the critical robustness issue for crash risk evaluation models and investigates countermeasures to enhance it. By mixing up crash and non-crash samples under the traffic flow fundamental diagram, traffic flow adversarial examples (TFAEs) were generated to simulate real-world traffic fluctuations. With the developed TF-AEs, model accuracy decreased by 8% and sensitivity dropped by 18%, indicating weak robustness of the baseline model (a convolutional neural network, CNN-based crash risk evaluation model). Then, a coverage-oriented adversarial training method was proposed to improve model robustness in highly imbalanced crash and non-crash situations and various crash risk transition patterns. Experiments showed that the proposed method was effective to improve model robustness as it could prevent 76.5% accuracy drops and 98.9% sensitivity drops against TF-AEs. Finally, the evaluation model outputs' stability and limitations of the current study are discussed.

关键词： Crash risk evaluation model model robustness Traffic flow fundamental diagram Traffic flow adversarial example Adversarial training

来源：评论

学校读者我要写书评

暂无评论

Improvement of empirical OPC model robustness using full-chip aerial image analysis

Improvement of empirical OPC model robustness using full-chi...

引用

23rd Annual BACUS Symposium on Photomask Technology

作者： Roessler, T Frankowsky, B Infineon Technol AG D-81609 Munich Germany

ISBN: (纸本)0819451436

With advanced CMOS technologies, model-based optical proximity correction (OPC) has become the most important aspect of post-tape-out data preparation for critical mask levels. While fabrication processes certainly remain the foundation of a qualified technology, the quality of OPC is increasingly moving into the focus of efforts to further improve yield. For a typical model-based OPC tool, the full OPC model consists of two distinct parts: (1) An aerial image part, based on a few, well-defined optical parameters of the lithography tool to describe the light intensity distribution in air at the wafer level and (2) an empirical part to model all other aspects of the pattern transfer, based on different black box modeling techniques such as kernel convolution or variable threshold modeling. Most importantly, the parameters for the empirical part are usually determined by fitting the model to proximity data measured from test structures. As a consequence, the robustness of the full OPC model for productive usage correlates directly with the extent to which these test structures provide a representative sampling of the configurations encountered in an actual product layout. In order to determine the quality of this sampling, full-chip aerial image analyses are performed for various mask levels of a product design. A comparison of the characteristics of the light intensity distributions of this design with the corresponding information obtained from the test structures reveals configurations that are not well covered by the latter. This insight allows the definition of suitable additional test structures in order to improve the robustness of subsequent empirical OPC models.

关键词： OPC Variable Threshold Resist models model robustness full-chip aerial image analysis OPC test patterns

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：