Upcycling adversarial attacks for infrared object detection

Kim, Hoseong; Lee, Chanyong

doi:10.1016/j.neucom.2022.01.090

Cited by 7 publications

(4 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For C&W, we chose [0.1, 1.0] as the range for c, and c is a parameter for box‐constraint in [27]. We have left all other parameters as defaults for implementing the recommendations [47].…”

Section: Experimental Results and Analysismentioning

confidence: 99%

DeepFake detection against adversarial examples based on D‐VAEGAN

Chen,

Xu,

2023

IET Image Processing

View full text Add to dashboard Cite

Recent years, the development of DeepFake has raise a lot of security problems. Therefore, detection of DeepFake is critical. However, the existing DeepFake detection methods are often vulnerable to adversarial attacks, i.e. adding carefully crafted imperceptible perturbations into forged images is possible to evade detection. In this paper, a DeepFake detection method based on image denoising is proposed by combining variational autoencoder (VAE) and generative adversarial network (GAN), namely D‐VAEGAN. Firstly, an encoder is designed to extract the features of the image in a low‐dimensional latent space. Then, a decoder reconstructs the original clean image using the features in this low‐dimensional latent space. Secondly, an auxiliary discriminative network is introduced to further improve the performance of the model, which improves the quality of the reconstructed images. Furthermore, feature similarity loss is added as a penalty term to the reconstruction optimization function to improve the adversarial robustness. Experimental results on the FaceForensics++ dataset show that the proposed approach significantly outperforms the five adversarial training‐based defence methods. The approach achieves 96% in accuracy, which is on average about 50% higher than other comparison methods.

show abstract

Section: Experimental Results and Analysismentioning

confidence: 99%

DeepFake detection against adversarial examples based on D‐VAEGAN

Chen,

Xu,

2023

IET Image Processing

View full text Add to dashboard Cite

show abstract

“…eps varies from 4/255 to 20/255 with step 4/255. The initial value

e p s = 4 false/ 255

is a default value used by the author of these seven attack algorithms 20,28–33 . We set the end value to

e p s = 20 false/ 255

.…”

Section: Discussionmentioning

confidence: 99%

“…Note that different domains (e.g., NLP and computer vision) require different adversarial attack algorithms. For example, in computer vision, the following algorithms can be representative: FGSM, 20 RFGSM, 28 FFGSM, 29 MIFGSM, 30 PGD, 31 TPGD, 32 and BIM 33 . Different attack algorithms can result in different attack performances.…”

Section: Methodsmentioning

confidence: 99%

Cheating your apps: Black‐box adversarial attacks on deep learning apps

Cao

Zhou

et al. 2023

J Software Evolu Process

View full text Add to dashboard Cite

Deep learning is a powerful technique to boost application performance in various fields, including face recognition, image classification, natural language understanding, and recommendation system. With the rapid increase in the computing power of mobile devices, developers can embed deep learning models into their apps for building more competitive products with more accurate and faster responses. Although there are several works of adversarial attacks against deep learning models in apps, they all need information about the models' internals (i.e., structures and weights) or need to modify the models. In this paper, we propose an effective black‐box approach by training substitute models to spoof the deep learning systems inside the apps. We evaluate our approach on 10 real‐world deep‐learning apps from Google Play to perform black‐box adversarial attacks. Through the study, we find three factors that can affect the performance of attacks. Our approach can reach a relatively high attack success rate of 66.60% on average. Compared with other adversarial attacks on mobile deep learning models, in terms of the average attack success rates, our approach outperforms its counterparts by 27.63%.

show abstract

“…Finally, we use the Auto attack [42], an adaptive attack that is widely utilized as a strong baseline thanks to its ability to overcome common defense mechanisms such as gradient masking [61]. All these attacks are implemented using a public library, Torchattacks [62]. The default setting in the library for other related parameters is applied for each attack.…”

Section: Evaluation Measuresmentioning

confidence: 99%

DRE: density-based data selection with entropy for adversarial-robust deep learning models

Guo

Cordy

et al. 2022

Neural Comput & Applic

View full text Add to dashboard Cite

Active learning helps software developers reduce the labeling cost when building high-quality machine learning models. A core component of active learning is the acquisition function that determines which data should be selected to annotate.State-of-the-art (SOTA) acquisition functions focus on clean performance (e.g. accuracy) but disregard robustness (an important quality property), leading to fragile models with negligible robustness (less than 0.20%). In this paper, we first propose to integrate adversarial training into active learning (adversarial-robust active learning, ARAL) to produce robust models. Our empirical study on 11 acquisition functions and 15105 trained deep neural networks (DNNs) shows that ARAL can produce models with robustness ranging from 2.35% to 63.85%. Our study also reveals, however, that the acquisition functions that perform well on accuracy are worse than random sampling when it comes to robustness. Via examining the reasons behind this, we devise the density-based robust sampling with entropy (DRE) to target both clean performance and robustness. The core idea of DRE is to maintain a balance between selected data and the entire set based on the entropy density distribution. DRE outperforms SOTA functions in terms of robustness by up to 24.40%, while remaining competitive on accuracy. Additionally, the in-depth evaluation shows that DRE is applicable as a test selection metric for model retraining and stands out from all compared functions by up to 8.21% robustness.

show abstract

Upcycling adversarial attacks for infrared object detection

Cited by 7 publications

References 4 publications

DeepFake detection against adversarial examples based on D‐VAEGAN

DeepFake detection against adversarial examples based on D‐VAEGAN

Cheating your apps: Black‐box adversarial attacks on deep learning apps

DRE: density-based data selection with entropy for adversarial-robust deep learning models

Contact Info

Product

Resources

About