mlsanityData to 7 Oct 2026

Adversarial Robustness in Multi-Task Learning: Promises and Illusions

arXiv 2110.15053 ↗First posted 26 Oct 2021cs.LG cs.AI cs.CVcited by 17

Not on the map. Why?

Abstract (arXiv)

Vulnerability to adversarial attacks is a well-known weakness of Deep Neural networks. While most of the studies focus on single-task neural networks with computer vision datasets, very little research has considered complex multi-task models that are common in real applications. In this paper, we evaluate the design choices that impact the robustness of multi-task deep learning networks. We provide evidence that blindly adding auxiliary tasks, or weighing the tasks provides a false sense of robustness. Thereby, we tone down the claim made by previous research and study the different factors which may affect robustness. In particular, we show that the choice of the task to incorporate in the loss function are important factors that can be leveraged to yield more robust models.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Builds on 21 of its 33 references that are in the map. First posted on arXiv 26 Oct 2021.

Builds on

21 of 33 references.

References · 21

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  3. 2013Intriguing properties of neural networkscited by 5,114
  4. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,427
  5. 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
  6. 2018BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learningcited by 1,282
  7. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,259
  8. 2019Habitat: A Platform for Embodied AI Researchcited by 1,163
  9. 2016Adversarial Machine Learning at Scalecited by 934
  10. 2017A Survey on Multi-Task Learningcited by 838
  11. 2017GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networkscited by 816
  12. 2017Evasion Attacks against Machine Learning at Test Timecited by 686
9 more
  1. 2018Taskonomy: Disentangling Task Transfer Learningcited by 631
  2. 2019Which Tasks Should Be Learned Together in Multi-task Learning?cited by 367
  3. 2017Certifying Some Distributional Robustness with Principled Adversarial Trainingcited by 333
  4. 2017On Detecting Adversarial Perturbationscited by 232
  5. 2017The Space of Transferable Adversarial Examplescited by 156
  6. 2017On the Robustness of Semantic Segmentation Models to Adversarial Attackscited by 126
  7. 2019Branched Multi-Task Networks: Deciding What Layers To Sharecited by 76
  8. 2018VLocNet++: Deep Multitask Learning for Semantic Visual Localization and Odometrycited by 61
  9. 2019Adversarial Embedding: A robust and elusive Steganography and Watermarking techniquecited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 1; 2022: 5; 2023: 3; 2024: 4; 2025: 2; 2026: 2.

17 papers.

Sort:

  1. 2023Sibling-Attack: Rethinking Transferable Adversarial Attacks against Face Recognitioncited by 32
  2. 2021A Unified Framework for Adversarial Attack and Defense in Constrained Feature Spacecited by 22
  3. 2022Detecting Adversarial Perturbations in Multi-Task Perceptioncited by 15
  4. 2024SafePowerGraph: Safety-aware Evaluation of Graph Neural Networks for Transmission Power Gridscited by 5
  5. 2023GAT: Guided Adversarial Training with Pareto-optimal Auxiliary Taskscited by 4
  6. 2022On Evaluating Adversarial Robustness of Chest X-ray Classification: Pitfalls and Best Practicescited by 3
  7. 2024A Multi-task Adversarial Attack Against Face Authenticationcited by 3
  8. 2022SkeleVision: Towards Adversarial Resiliency of Person Tracking with Multi-Task Learningcited by 2
  9. 2024Adversarial Attacks on Hidden Tasks in Multi-Task Learningcited by 2
  10. 2025Merge Now, Regret Later: The Hidden Cost of Model Merging Is Adversarial Transferabilitycited by 2
  11. 2022Multitasking Models are Robust to Structural Failure: A Neural Model for Bilingual Cognitive Reservecited by 1
  12. 2023Attacking All Tasks at Once Using Adversarial Examples in Multi-Task Learningcited by 1
5 more
  1. 2024Stealthy Multi-Task Adversarial Attackscited by 1
  2. 2025Multi-task Adversarial Attacks against Black-box Model with Few-shot Queriescited by 1
  3. 2026RobuMTL: Enhancing Multi-Task Learning Robustness Against Weather Conditionscited by 1
  4. 2022Hear No Evil: Towards Adversarial Robustness of Automatic Speech Recognition via Multi-Task Learningcited by 0
  5. 2026Multi-Task Consistency-based Detection of Adversarial Attackscited by 0