|
Аннотация.
В статье рассмотрена актуальная проблема уязвимости технологий искусственного интеллекта на основе нейронных сетей в задаче распознавания образов. Показано, что применение нейронных сетей порождает множество уязвимостей. Приведены конкретные примеры таких уязвимостей: некорректная классификация изображений, содержащих вредоносный шум или заплатки, отказ распознающих систем при наличии на изображении особых узоров, в том числе нанесенных на объекты реального мира, отравление обучающей выборки и др. На основе проведенного анализа показана необходимость улучшения безопасности технологий искусственного интеллекта и даны предложения, способствующие этому улучшению.
Ключевые слова:
нейронные сети, атаки на нейронные сети, вредоносные изображения, нейросетевая безопасность.
Стр. 49-58.
DOI 10.14357/20718632230405
EDN BVJZWS Литература
1. Ye M. et al. Deep learning for person re-identification: A survey and outlook //IEEE transactions on pattern analysis and machine intelligence. – 2021. – Т. 44. – №. 6. – С. 2872-2893. 2. Arlazarov V. V., Andreeva E. I., Bulatov K. B., Nikolaev D. P., Petrova O. O., Savelev B. I., Slavin O. A. Document image analysis and recognition: A survey // Компьютерная оптика. — 2022. — Т. 46. — № 4. — С. 567-589 3. Yang B. et al. Edge intelligence for autonomous driving in 6G wireless system: Design challenges and solutions //IEEE Wireless Communications. – 2021. – Т. 28. – №. 2. – С. 40-47. 4. Gu T., Dolan-Gavitt B., Garg S. Badnets: Identifying vulnerabilities in the machine learning model supply chain //arXiv preprint arXiv:1708.06733. – 2017. 5. Fredrikson M., Jha S., Ristenpart T. Model inversion attacks that exploit confidence information and basic countermeasures //Proceedings of the 22nd ACM SIGSAC conference on computer and communications security. – 2015. – С. 1322-1333. 6. Szegedy C. et al. Intriguing properties of neural networks //arXiv preprint arXiv:1312.6199. – 2013. 7. Brown T. B. et al. Adversarial patch //arXiv preprint arXiv:1712.09665. – 2017. 8. Lin C. S. et al. Real-world adversarial examples via makeup //ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). – IEEE, 2022. – С. 2854-2858. 9. Hu S. et al. Protecting facial privacy: Generating adversarial identity masks via style-robust makeup transfer //Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. – 2022. – С. 15014-15023. 10. Zolfi A. et al. Adversarial Mask: Real-World Universal Adversarial Attack on Face Recognition Models //Joint European Conference on Machine Learning and Knowledge Discovery in Databases. – Cham : Springer Nature Switzerland, 2022. – С. 304-320. 11. Zhou Z. et al. Invisible mask: Practical attacks on face recognition with infrared //arXiv preprint arXiv:1803.04683. – 2018. 12. Wu Z., Lim S.N., Davis L.S., Goldstein T. Making an invisibility cloak: Real world adversarial attacks on object detectors. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16 2020 (pp. 1-17). Springer International Publishing. 13. Thys S., Van Ranst W., Goedemé T. Fooling automated surveillance cameras: adversarial patches to attack person detection //Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. – 2019. – С. 0-0. 14. Chen J. et al. Diffusion Models for Imperceptible and Transferable Adversarial Attack //arXiv preprint arXiv:2305.08192. – 2023. 15. Hong S. et al. Security analysis of deep neural networks operating in the presence of cache side-channel attacks //arXiv preprint arXiv:1810.03487. – 2018. 16. Oh S. J., Schiele B., Fritz M. Towards reverse-engineering black-box neural networks //Explainable AI: Interpreting, Explaining and Visualizing Deep Learning. – 2019. – С. 121-144. 17. Chmielewski Ł., Weissbart L. On reverse engineering neural network implementation on gpu //Applied Cryptography and Network Security Workshops: ACNS 2021 Satellite Workshops, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S&P, SCI, SecMT, and SiMLA, Kamakura, Japan, June 21–24, 2021, Proceedings. – Springer International Publishing, 2021. – С. 96-113. 18. Goldblum M. et al. Dataset security for machine learning: Data poisoning, backdoor attacks, and defenses //IEEE Transactions on Pattern Analysis and Machine Intelligence. – 2022. – Т. 45. – №. 2. – С. 1563-1580. 19. Shafahi A. et al. Poison frogs! targeted clean-label poisoning attacks on neural networks //Advances in neural information processing systems. – 2018. – Т. 31. 20. Wang Y. et al. Sapag: A self-adaptive privacy attack from gradients //arXiv preprint arXiv:2009.06228. – 2020. 21. Warr K. Strengthening deep neural networks: Making AI less susceptible to adversarial trickery. – O'Reilly Media, 2019. 22. Long T. et al. A survey on adversarial attacks in computer vision: Taxonomy, visualization and future directions //Computers & Security. – 2022. – С. 102847. 23. Akhtar N., Mian A. Threat of adversarial attacks on deep learning in computer vision: A survey //Ieee Access. – 2018. – Т. 6. – С. 14410-14430 24. Machado G. R., Silva E., Goldschmidt R. R. Adversarial machine learning in image classification: A survey toward the defender’s perspective //ACM Computing Surveys (CSUR). – 2021. – Т. 55. – №. 1. – С. 1-38. 25. Ren K. et al. Adversarial attacks and defenses in deep learning //Engineering. – 2020. – Т. 6. – №. 3. – С. 346-360. 26. Zhang X. et al. Imperceptible black-box waveform-level adversarial attack towards automatic speaker recognition //Complex & Intelligent Systems. – 2023. – Т. 9. – №. 1. – С. 65-79. 27. Kwon H., Lee S. Ensemble transfer attack targeting text classification systems //Computers & Security. – 2022. – Т. 117. – С. 102695. 28. Mo K. et al. Attacking deep reinforcement learning with decoupled adversarial policy //IEEE Transactions on Dependable and Secure Computing. – 2022. – Т. 20. – №.1. – С. 758-768. 29. Zhou X. et al. Hierarchical adversarial attacks against graph-neural-network-based IoT network intrusion detection system //IEEE Internet of Things Journal. – 2021. – Т. 9. – №. 12. – С. 9310-9319. 30. Kumar R. S. S. et al. Adversarial machine learning-industry perspectives //2020 IEEE Security and Privacy Workshops (SPW). – IEEE, 2020. – С. 69-75. 31. Paleyes A., Urma R. G., Lawrence N. D. Challenges in deploying machine learning: a survey of case studies //ACM Computing Surveys. – 2022. – Т. 55. – №. 6. – С. 1-29. 32. Ala-Pietilä P. et al. The assessment list for trustworthy artificial intelligence (ALTAI). – European Commission, 2020. 33. Musser M. et al. Adversarial Machine Learning and Cybersecurity: Risks, Challenges, and Legal Implications //arXiv preprint arXiv:2305.14553. – 2023. 34. Facial recognition’s latest foe: Italian knitwear [Электронный ресурс] // The Record. URL: https://therecord.media/facialrecognitions-latest-foe-italian-knitwear (дата обращения: 20.07.2023). 35. Как мы боремся с копированием контента, или первая adversarial attack в проде. [Электронный ресурс] // Хабр. URL: https://habr.com/ru/companies/avito/articles/452142 (дата обращения: 20.07.2023). 36. Povolny S., Trivedi S. Model hacking ADAS to pave safer roads for autonomous vehicles. [Электронный ресурс] // McAfee Blogs. URL: https://www.mcafee.com/blogs/otherblogs/mcafee-labs/model-hacking-adas-to-pave-safer-roadsfor-autonomous-vehicles/ (дата обращения: 20.07.2023).
|
