Optimization-based adversarial perturbations against twin support vector machines

Article Computer Science, Interdisciplinary Applications

Robust and Distributionally Robust Optimization Models for Linear Support Vector Machine

Daniel Faccini et al.

Summary: This paper presents novel data-driven optimization models for improving the classification performance of Support Vector Machines (SVM). By introducing uncertainty sets and robust optimization models, more reliable classification can be achieved in real-life noisy data. Experimental results show that this method is particularly beneficial for data sets with a small number of observations and can improve out-of-sample accuracy as the dimension of the data sets increases.

COMPUTERS & OPERATIONS RESEARCH (2022)

Add to Collection

Article Computer Science, Theory & Methods

Fast Locally Optimal Detection of Targeted Universal Adversarial Perturbations

Amish Goel et al.

Summary: This paper proposes a locally-optimal generalized likelihood ratio test (LO-GLRT) for detecting targeted attacks on a classifier, which involves adding a norm-bounded targeted universal adversarial perturbation (UAP) to the classifier's input. The paper analyzes and evaluates the test, finding it to be effective in detecting the attacks.

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

Universal adversarial perturbations against object detection

Debang Li et al.

Summary: This paper presents an algorithm to generate universal adversarial perturbations in object detection, which are visually imperceptible but can cause state-of-the-art detectors to fail. Compared to image-specific perturbations, universal perturbations are more challenging and pose a greater threat to AI security.

PATTERN RECOGNITION (2021)

Add to Collection

Article Computer Science, Interdisciplinary Applications

New method for solving Ivanov regularization-based support vector machine learning

Xiang Xu et al.

Summary: The SVM model, based on the SRM principle, including I-SVM, is well-known in machine learning. A new method for optimizing I-SVM was proposed in this paper, showing improved performance through experimental results.

COMPUTERS & OPERATIONS RESEARCH (2021)

Add to Collection

Article Computer Science, Information Systems

A novel model for anomaly detection in network traffic based on kernel support vector machine

Qian Ma et al.

Summary: Machine learning models, specifically the SVM-L, are proposed for anomaly detection in network traffic by treating raw URLs as natural language and transforming them into mathematical vectors. The model achieves over 99% accuracy and outperforms state-of-the-art methods in standard evaluation measurements.

COMPUTERS & SECURITY (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

Fooling deep neural detection networks with adaptive object-oriented adversarial perturbation

Yatie Xiao et al.

Summary: Deep learning excels at complex tasks, but Deep Neural Networks are vulnerable to carefully crafted adversarial perturbations. The AO(2)AM algorithm focuses on object-level adversarial perturbations to fool deep neural object detection networks effectively.

PATTERN RECOGNITION (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Data-free Universal Adversarial Perturbation and Black-box Attack

Chaoning Zhang et al.

Summary: Our research focuses on providing an alternative explanation for the phenomenon of untargeted UAP, aiming to reduce dependence on original training samples and exploring the potential for data-free black-box attacks. Our work proposes utilizing artificial Jigsaw images as training samples for competitive performance in universal adversarial perturbations.

2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021) (2021)

Add to Collection

Article Mathematics, Applied