find papers on evaluation metrics of adversarial attacks on face recognition

Papers on Evaluation Metrics of Adversarial Attacks on Face Recognition

Cross-resolution face recognition adversarial attacks
- Summary: This paper explores the impact of adversarial attacks on face recognition systems in cross-resolution scenarios. It evaluates the performance of face recognition models against adversarial attacks using L1, L2, and L∞ metrics. The study finds that deep representation attacks pose a significant threat and that image resolution affects the success of adversarial attacks.
- Citation:
Attacks on state-of-the-art face recognition using attentional adversarial attack generative network
- Summary: This research introduces the Attentional Adversarial Attack Generative Network (A3GN) to generate adversarial examples that mislead face recognition systems. The paper evaluates the effectiveness of these attacks in evading recognition by state-of-the-art networks.
- Citation:
Simultaneous Adversarial Attacks On Multiple Face Recognition System Components
- Summary: This study investigates adversarial attacks targeting multiple components of face recognition systems, such as the face detector and feature extractor. The paper demonstrates the effectiveness of multi-objective attacks and suggests ways to improve robustness by leveraging attack vector knowledge.
- Citation:
Delving into the Adversarial Robustness on Face Recognition
- Summary: This paper performs comprehensive robustness evaluations of 15 popular face recognition models against various adversarial attacks, including dodging and impersonation attacks, and L2 and L∞ attacks. It proposes a landmark-guided cutout (LGC) attack method to improve the transferability of adversarial examples.
- Citation:
Meaningful Adversarial Stickers for Face Recognition in Physical World
- Summary: The paper proposes a method using meaningful adversarial stickers to attack face recognition systems in a black-box setting. It evaluates the attack success rates on different models and demonstrates the method's effectiveness in complex physical conditions.
- Citation:
An efficient face recognition attack method based on generative adversarial networks and cosine metrics
- Summary: This research introduces an attack method using generative adversarial networks (HNUGAN) and cosine metrics to create adversarial samples. The paper evaluates the impact of these samples on the recognition accuracy of models like GoogleNet and ResNet.
- Citation:
Adversarial Attacks on Convolutional Neural Networks in Facial Recognition Domain
- Summary: The paper explores various adversarial attack algorithms on facial image datasets, including single-pixel and all-pixel attacks. It evaluates the robustness of deep neural networks in the facial recognition domain and the effectiveness of different attack methods.
- Citation:
Adversarial Attacks Against Face Recognition: A Comprehensive Study
- Summary: This comprehensive survey reviews adversarial attacks against face recognition systems and evaluates the effectiveness of different attack and defense methods. It proposes a taxonomy based on various criteria and discusses challenges and future research directions.
- Citation:
Detection of Face Recognition Adversarial Attacks
- Summary: The paper presents a detection approach for adversarial samples in face recognition systems. It evaluates the effectiveness of the approach against classifier attacks and deep features attacks, demonstrating its generalizability to different types of offensives.
- Citation:
Adv-Attribute: Inconspicuous and Transferable Adversarial Attack on Face Recognition
- Summary: This research proposes the Adv-Attribute framework to generate inconspicuous and transferable adversarial attacks on face recognition by perturbing high-level semantics. The paper evaluates the attack success rates and visual effects against recent attack methods.
- Citation:

Find papers on evaluation metrics of adversarial attacks on face recognition

Summary

Papers on Evaluation Metrics of Adversarial Attacks on Face Recognition

Sources and full results

Cross-resolution face recognition adversarial attacks

Deep representation attacks are more dangerous for face recognition systems than classification output attacks, and cross-resolution training approaches can improve their resilience to adversarial attacks.

Attacks on state-of-the-art face recognition using attentional adversarial attack generative network

The Attentional Adversarial Attack Generative Network (A 3 G N) effectively generates adversarial examples that can evade recognition by state-of-the-art face recognition networks and most of them are recognized as the target person.

Simultaneous Adversarial Attacks On Multiple Face Recognition System Components

Adversarial attacks can effectively fool multiple face recognition system components, potentially reducing face detection probability by up to 50%.

Delving into the Adversarial Robustness on Face Recognition

This paper evaluates 15 popular face recognition models' robustness against various adversarial attacks, providing insights for future research on face recognition.

Meaningful Adversarial Stickers for Face Recognition in Physical World

Meaningful Adversarial Stickers effectively attack face recognition systems, achieving success rates of 81.78%, 72.93%, and 79.26% in complex physical conditions.

An efficient face recognition attack method based on generative adversarial networks and cosine metrics

This study proposes an efficient face recognition attack method using generative adversarial networks and cosine metrics, reducing recognition accuracy of models like GoogleNet and ResNet to a very low level.

Adversarial Attacks on Convolutional Neural Networks in Facial Recognition Domain

Adversarial attacks on Deep Neural Networks can significantly decrease confidence in facial recognition, but face images remain clearly recognizable to a human.

Adversarial Attacks Against Face Recognition: A Comprehensive Study

Adversarial attacks against face recognition systems can lead to incorrect output predictions, and this paper presents a taxonomy of existing attack and defense methods to categorize them for further research.

Detection of Face Recognition Adversarial Attacks

Our adversarial detection approach effectively detects adversarial samples against Face Recognition systems, improving forensics and demonstrating its generalizability to different types of offensives.

Adv-Attribute: Inconspicuous and Transferable Adversarial Attack on Face Recognition

The Adv-Attribute framework effectively generates inconspicuous and transferable attacks on face recognition by perturbing high-level semantics, improving attack transferability and reducing vulnerability to denoising defense models.

Try another search