A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference

• URL: http://arxiv.org/abs/2010.02432v2
• Date: Thu, 25 Feb 2021 22:38:35 GMT
• Title: A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference
• Authors: Sanghyun Hong, Yi\u{g}itcan Kaya, Ionu\c{t}-Vlad Modoranu, Tudor Dumitra\c{s}
• Abstract summary: A slowdown attack reduces the efficacy of multi-exit DNNs by 90-100%, and it amplifies the latency by 1.5-5$times$ in a typical IoT deployment. We show that it is possible to craft universal, reusable perturbations and that the attack can be effective in realistic black-box scenarios.
• Score: 6.320009081099895
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent increases in the computational demands of deep neural networks (DNNs), combined with the observation that most input samples require only simple models, have sparked interest in $input$-$adaptive$ multi-exit architectures, such as MSDNets or Shallow-Deep Networks. These architectures enable faster inferences and could bring DNNs to low-power devices, e.g., in the Internet of Things (IoT). However, it is unknown if the computational savings provided by this approach are robust against adversarial pressure. In particular, an adversary may aim to slowdown adaptive DNNs by increasing their average inference time$-$a threat analogous to the $denial$-$of$-$service$ attacks from the Internet. In this paper, we conduct a systematic evaluation of this threat by experimenting with three generic multi-exit DNNs (based on VGG16, MobileNet, and ResNet56) and a custom multi-exit architecture, on two popular image classification benchmarks (CIFAR-10 and Tiny ImageNet). To this end, we show that adversarial example-crafting techniques can be modified to cause slowdown, and we propose a metric for comparing their impact on different architectures. We show that a slowdown attack reduces the efficacy of multi-exit DNNs by 90-100%, and it amplifies the latency by 1.5-5$\times$ in a typical IoT deployment. We also show that it is possible to craft universal, reusable perturbations and that the attack can be effective in realistic black-box scenarios, where the attacker has limited knowledge about the victim. Finally, we show that adversarial training provides limited protection against slowdowns. These results suggest that further research is needed for defending multi-exit architectures against this emerging threat. Our code is available at https://github.com/sanghyun-hong/deepsloth.

Related papers

• A Geometrical Approach to Evaluate the Adversarial Robustness of Deep Neural Networks [52.09243852066406]
  Adversarial Converging Time Score (ACTS) measures the converging time as an adversarial robustness metric. We validate the effectiveness and generalization of the proposed ACTS metric against different adversarial attacks on the large-scale ImageNet dataset.
  arXiv  Detail & Related papers  (2023-10-10T09:39:38Z)
• Not So Robust After All: Evaluating the Robustness of Deep Neural Networks to Unseen Adversarial Attacks [5.024667090792856]
  Deep neural networks (DNNs) have gained prominence in various applications, such as classification, recognition, and prediction. A fundamental attribute of traditional DNNs is their vulnerability to modifications in input data, which has resulted in the investigation of adversarial attacks. This study aims to challenge the efficacy and generalization of contemporary defense mechanisms against adversarial attacks.
  arXiv  Detail & Related papers  (2023-08-12T05:21:34Z)
• Quantization Aware Attack: Enhancing Transferable Adversarial Attacks by Model Quantization [57.87950229651958]
  Quantized neural networks (QNNs) have received increasing attention in resource-constrained scenarios due to their exceptional generalizability. Previous studies claim that transferability is difficult to achieve across QNNs with different bitwidths. We propose textitquantization aware attack (QAA) which fine-tunes a QNN substitute model with a multiple-bitwidth training objective.
  arXiv  Detail & Related papers  (2023-05-10T03:46:53Z)
• Training High-Performance Low-Latency Spiking Neural Networks by Differentiation on Spike Representation [70.75043144299168]
  Spiking Neural Network (SNN) is a promising energy-efficient AI model when implemented on neuromorphic hardware. It is a challenge to efficiently train SNNs due to their non-differentiability. We propose the Differentiation on Spike Representation (DSR) method, which could achieve high performance.
  arXiv  Detail & Related papers  (2022-05-01T12:44:49Z)
• Robustness of Bayesian Neural Networks to White-Box Adversarial Attacks [55.531896312724555]
  Bayesian Networks (BNNs) are robust and adept at handling adversarial attacks by incorporating randomness. We create our BNN model, called BNN-DenseNet, by fusing Bayesian inference (i.e., variational Bayes) to the DenseNet architecture. An adversarially-trained BNN outperforms its non-Bayesian, adversarially-trained counterpart in most experiments.
  arXiv  Detail & Related papers  (2021-11-16T16:14:44Z)
• Detect and Defense Against Adversarial Examples in Deep Learning using Natural Scene Statistics and Adaptive Denoising [12.378017309516965]
  We propose a framework for defending DNN against ad-versarial samples. The detector aims to detect AEs bycharacterizing them through the use of natural scenestatistic. The proposed method outperforms the state-of-the-art defense techniques.
  arXiv  Detail & Related papers  (2021-07-12T23:45:44Z)
• BreakingBED -- Breaking Binary and Efficient Deep Neural Networks by Adversarial Attacks [65.2021953284622]
  We study robustness of CNNs against white-box and black-box adversarial attacks. Results are shown for distilled CNNs, agent-based state-of-the-art pruned models, and binarized neural networks.
  arXiv  Detail & Related papers  (2021-03-14T20:43:19Z)
• Adversarial Attacks on Deep Learning Based Power Allocation in a Massive MIMO Network [62.77129284830945]
  We show that adversarial attacks can break DL-based power allocation in the downlink of a massive multiple-input-multiple-output (maMIMO) network. We benchmark the performance of these attacks and show that with a small perturbation in the input of the neural network (NN), the white-box attacks can result in infeasible solutions up to 86%.
  arXiv  Detail & Related papers  (2021-01-28T16:18:19Z)
• Towards Universal Physical Attacks On Cascaded Camera-Lidar 3D Object Detection Models [16.7400223249581]
  We propose a universal and physically realizable adversarial attack on a cascaded multi-modal deep learning network (DNN) We show that the proposed universal multi-modal attack was successful in reducing the model's ability to detect a car by nearly 73%.
  arXiv  Detail & Related papers  (2021-01-26T12:40:34Z)
• Progressive Defense Against Adversarial Attacks for Deep Learning as a Service in Internet of Things [9.753864027359521]
  Some Deep Neural Networks (DNN) can be easily misled by adding relatively small but adversarial perturbations to the input. We present a defense strategy called a progressive defense against adversarial attacks (PDAAA) for efficiently and effectively filtering out the adversarial pixel mutations. The result shows it outperforms the state-of-the-art while reducing the cost of model training by 50% on average.
  arXiv  Detail & Related papers  (2020-10-15T06:40:53Z)
• DeepHammer: Depleting the Intelligence of Deep Neural Networks through Targeted Chain of Bit Flips [29.34622626909906]
  We demonstrate the first hardware-based attack on quantized deep neural networks (DNNs) DeepHammer is able to successfully tamper DNN inference behavior at run-time within a few minutes. Our work highlights the need to incorporate security mechanisms in future deep learning system.
  arXiv  Detail & Related papers  (2020-03-30T18:51:59Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.