Generative Dynamic Patch Attack

Li, Xiang; Ji, Shihao

doi:10.48550/arxiv.2111.04266

Cited by 4 publications

(3 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, Su et al discuss [17] a one-pixel attack technique that is another valuable subset of patch attacks based on evolutionary strategies. Regarding dynamic patch generation, Li and Ji [18] present a flexible model capable of producing visible or non-visible patches that can move around an image to find a suitable attack position. In another investigation, Mohapatra et al [19] discuss semantic perturbation over discrete and continuous parameters that allow the model to predict based on dimensional perturbations.…”

Section: Related Workmentioning

confidence: 99%

The Impact of Simultaneous Adversarial Attacks on Robustness of Medical Image Analysis

Pal,

Rahman,

Beheshti

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Deep learning models are widely used in healthcare systems. However, deep learning models are vulnerable to attacks themselves. Significantly, due to the black-box nature of the deep learning model, it is challenging to detect attacks. Furthermore, due to data sensitivity, such adversarial attacks in healthcare systems are considered potential security and privacy threats. In this paper, we provide a comprehensive analysis of adversarial attacks on medical image analysis, including two adversary methods, FGSM and PGD, applied to an entire image or partial image. The partial attack comes in various sizes, either the individual or combinational format of attack. We use three medical datasets to examine the impact of the model's accuracy and robustness. Finally, we provide a complete implementation of the attacks and discuss the results. Our results indicate the weakness and robustness of four deep learning models and exhibit how varying perturbations stimulate model behaviour regarding the specific area and critical features.

show abstract

Section: Related Workmentioning

confidence: 99%

The Impact of Simultaneous Adversarial Attacks on Robustness of Medical Image Analysis

Pal,

Rahman,

Beheshti

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The first physical attack against deep neural networks was proposed by [7], by developing an algorithm for printing adversarial eyeglass frames able to evade a face recognition system. Brown et al [5] introduced the first universal patch attack that Attack Cross-model Transfer Targeted Untargeted Transformations Sharif et al [7] rot Brown et al [5] loc, scl, rot LaVAN [8] loc PS-GAN [25] loc DT-Patch [26] PatchAttack [27] loc, scl IAPA [28] Lennon et al [29] loc, scl, rot Xiao et al [30] various Ye et al [31] loc, scl, rot GDPA [32] loc Ours loc, rot…”

Section: Patch Attacksmentioning

confidence: 99%

“…Ye et al [31] study the specific application of patch attacks on traffic sign recognition and use an ensemble of models to improve the attack success rate. The Generative Dynamic Patch Attack (GDPA), proposed by Li et al [32], generates the patch pattern and location for each input image simultaneously, reducing the runtime of the attack and making it hence a good candidate to use for adversarial training.…”

Section: Patch Attacksmentioning

confidence: 99%

ImageNet-Patch: A dataset for benchmarking machine learning robustness against adversarial patches

Pintor

Angioni²,

Sotgiu³

et al. 2023

Pattern Recognition

View full text Add to dashboard Cite

The Path to Defence: A Roadmap to Characterising Data Poisoning Attacks on Victim Models

Chaalan,

Pang,

Kamruzzaman

et al. 2024

ACM Comput. Surv.

View full text Add to dashboard Cite

The approach and process of Data Poisoning Attacks (DPA) to distort training data to machine learning model and manipulate the model behaviours is not only technically complex but also often victim model dependent. To protect the victim model, the vast number of DPAs and their variants make defenders rely on trial and error techniques to find the ultimate defence solution which is exhausting and very time-consuming. This paper comprehensively summarises the latest research on DPAs and defences, proposes a DPA characterizing model to help investigate adversary attacks dependency on the victim model, and builds a DPA roadmap as the path navigating to defence. Having the roadmap as an applied framework that contains DPA families sharing the same features and mathematical computations will equip the defenders with a powerful tool to quickly find the ultimate defences, away from the exhausting trial and error methodology. The roadmap validated by use cases has been made available as an open access platform, enabling other researchers to add in new DPAs and update the map continuously.

show abstract

Generative Dynamic Patch Attack

Cited by 4 publications

References 21 publications

The Impact of Simultaneous Adversarial Attacks on Robustness of Medical Image Analysis

The Impact of Simultaneous Adversarial Attacks on Robustness of Medical Image Analysis

ImageNet-Patch: A dataset for benchmarking machine learning robustness against adversarial patches

The Path to Defence: A Roadmap to Characterising Data Poisoning Attacks on Victim Models

Contact Info

Product

Resources

About