Papers
Topics
Authors
Recent
Search
2000 character limit reached

Federated Learning Priorities Under the European Union Artificial Intelligence Act

Published 5 Feb 2024 in cs.LG, cs.AI, cs.CY, and cs.DC | (2402.05968v1)

Abstract: The age of AI regulation is upon us, with the European Union Artificial Intelligence Act (AI Act) leading the way. Our key inquiry is how this will affect Federated Learning (FL), whose starting point of prioritizing data privacy while performing ML fundamentally differs from that of centralized learning. We believe the AI Act and future regulations could be the missing catalyst that pushes FL toward mainstream adoption. However, this can only occur if the FL community reprioritizes its research focus. In our position paper, we perform a first-of-its-kind interdisciplinary analysis (legal and ML) of the impact the AI Act may have on FL and make a series of observations supporting our primary position through quantitative and qualitative analysis. We explore data governance issues and the concern for privacy. We establish new challenges regarding performance and energy efficiency within lifecycle monitoring. Taken together, our analysis suggests there is a sizable opportunity for FL to become a crucial component of AI Act-compliant ML systems and for the new regulation to drive the adoption of FL techniques in general. Most noteworthy are the opportunities to defend against data bias and enhance private and secure computation

Definition Search Book Streamline Icon: https://streamlinehq.com
References (78)
  1. Optimal allocation of the EU carbon budget: A multi-model assessment, may 2023. URL https://ariadneprojekt.de/media/2022/06/Ariadne-Analysis_Carbon-Budget-multi-model-assessment_June2022.pdf. Document 32023L0959.
  2. Green Federated Learning via Energy-Aware Client Selection. In GLOBECOM 2022 - 2022 IEEE Global Communications Conference, pp.  13–18, 2022. doi: 10.1109/GLOBECOM48099.2022.10001569.
  3. Differentially private learning with adaptive clipping. Advances in Neural Information Processing Systems, 34:17455–17466, 2021.
  4. Leveraging NVLINK and asynchronous data transfer to scale beyond the memory capacity of GPUs. In Proceedings of the 8th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, SC ’17. ACM, November 2017. doi: 10.1145/3148226.3148232. URL http://dx.doi.org/10.1145/3148226.3148232.
  5. SLoRA: Federated parameter efficient fine-tuning of language models, 2023. URL https://arxiv.org/abs/2308.06522.
  6. Flower: A Friendly Federated Learning Research Framework, 2020. URL https://arxiv.org/abs/2007.14390.
  7. Practical Secure Aggregation for Privacy-Preserving Machine Learning. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, CCS ’17. ACM, October 2017. doi: 10.1145/3133956.3133982. URL http://dx.doi.org/10.1145/3133956.3133982.
  8. Butskoy, D. Linux Traceroute, 12 2023. URL http://traceroute.sourceforge.net/.
  9. pFL-Bench: A Comprehensive Benchmark for Personalized Federated Learning. In Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2022. URL https://openreview.net/forum?id=2ptbv_JjYKA.
  10. CMS Law. GDPR Enforcement Tracker, 01 2024. URL https://www.enforcementtracker.com/.
  11. A Survey of Man In The Middle Attacks. IEEE Communications Surveys & Tutorials, 18(3):2027–2051, 2016. doi: 10.1109/COMST.2016.2548426.
  12. Council of the European Union. Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL - LAYING DOWN HARMONISED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND AMENDING CERTAIN UNION LEGISLATIVE ACTS, apr 2021. URL https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52021PC0206. Document 52021PC0206.
  13. Trends in AI inference energy consumption: Beyond the performance-vs-parameter laws of deep learning. Sustainable Computing: Informatics and Systems, 38:100857, April 2023. ISSN 2210-5379. doi: 10.1016/j.suscom.2023.100857. URL http://dx.doi.org/10.1016/j.suscom.2023.100857.
  14. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, 2018. URL https://arxiv.org/abs/1810.04805.
  15. The Algorithmic Foundations of Differential Privacy. Foundations and Trends® in Theoretical Computer Science, 9(3–4):211–407, 2013. ISSN 1551-3068. doi: 10.1561/0400000042. URL http://dx.doi.org/10.1561/0400000042.
  16. Federated learning for water consumption forecasting in smart cities. In ICC 2023-IEEE International Conference On Communications, pp.  1798–1803. IEEE, 2023.
  17. European Commission. Market analysis - Electricity market - recent developments, 07 2023a. URL https://energy.ec.europa.eu/data-and-analysis/market-analysis_en.
  18. European Commission. A European approach to artificial intelligence, 2023b. URL https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence.
  19. European Environment Agency. Greenhouse gas emission intensity of electricity generation, Oct 2023. URL https://www.eea.europa.eu/data-and-maps/daviz/co2-emission-intensity-14#tab-chart_7.
  20. European Parliament and Council. Laying down harmonised rules on artificial intelligence (artificial intelligence act) and amending certain union legislative acts, 01 2024. URL https://www.linkedin.com/posts/dr-laura-caroli-0a96a8a_ai-act-consolidated-version-activity-7155181240751374336-B3Ym/.
  21. Eurostat. Electricity prices for household consumers, Oct 2023. URL https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Electricity_price_statistics.
  22. Human behaviour as an aspect of cybersecurity assurance. Security and Communication Networks, 9(17):4667–4679, 2016.
  23. Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach. In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., and Lin, H. (eds.), Advances in Neural Information Processing Systems, volume 33, pp.  3557–3568. Curran Associates, Inc., 2020. URL https://proceedings.neurips.cc/paper_files/paper/2020/file/24389bfe4fe2eba8bf9aa9203a44cdad-Paper.pdf.
  24. The pile: An 800gb dataset of diverse text for language modeling. arXiv preprint arXiv:2101.00027, 2020.
  25. Inverting Gradients - How easy is it to break privacy in federated learning? In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., and Lin, H. (eds.), Advances in Neural Information Processing Systems, volume 33, pp.  16937–16947. Curran Associates, Inc., 2020. URL https://proceedings.neurips.cc/paper_files/paper/2020/file/c4ede56bbd98819ae6112b20ac6bf145-Paper.pdf.
  26. Google. 24/7 Carbon-Free Energy by 2030, 2023. URL https://www.google.com/about/datacenters/cleanenergy/.
  27. Amnesiac machine learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pp.  11516–11524, 2021.
  28. Deep Learning with Limited Numerical Precision. In Bach, F. R. and Blei, D. M. (eds.), Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015, volume 37 of JMLR Workshop and Conference Proceedings, pp.  1737–1746. JMLR.org, 2015. URL http://proceedings.mlr.press/v37/gupta15.html.
  29. Federated Unlearning: How to Efficiently Erase a Client in FL?, 2022. URL https://arxiv.org/abs/2207.05521.
  30. FedMLSecurity: A Benchmark for Attacks and Defenses in Federated Learning and Federated LLMs, 2023.
  31. FedML: A Research Library and Benchmark for Federated Machine Learning, 2020. URL https://arxiv.org/abs/2007.13518.
  32. House Of Commons of Canada. An Act to enact the Consumer Privacy Protection Act, the Personal Information and Data Protection Tribunal Act and the Artificial Intelligence and Data Act and to make consequential and related amendments to other Acts, 6 2022. URL https://www.parl.ca/DocumentViewer/en/44-1/bill/C-27/first-reading.
  33. Rethinking Federated Learning With Domain Shift: A Prototype View. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  16312–16322, June 2023.
  34. Evaluating Gradient Inversion Attacks and Defenses in Federated Learning. In Beygelzimer, A., Dauphin, Y., Liang, P., and Vaughan, J. W. (eds.), Advances in Neural Information Processing Systems, 2021. URL https://openreview.net/forum?id=0CDKgyYaxC8.
  35. Energy Consumption of Content Distribution from Nano Data Centers versus Centralized Data Centers. ACM SIGMETRICS Performance Evaluation Review, 42(3):49–54, December 2014. ISSN 0163-5999. doi: 10.1145/2695533.2695555. URL http://dx.doi.org/10.1145/2695533.2695555.
  36. FedML-HE: An Efficient Homomorphic-Encryption-Based Privacy-Preserving Federated Learning System, 2023. URL https://arxiv.org/abs/2303.10837.
  37. Green, Quantized Federated Learning over Wireless Networks: An Energy-Efficient Design. IEEE Transactions on Wireless Communications, pp.  1–1, 2023. doi: 10.1109/TWC.2023.3289177.
  38. The law of recitals in European Community legislation. ILSA J. Int’l & Comp. L., 15:61, 2008.
  39. Towards Unbounded Machine Unlearning. arXiv preprint arXiv:2302.09880, 2023.
  40. Lang, K. NewsWeeder: Learning to Filter Netnews. In Prieditis, A. and Russell, S. (eds.), Machine Learning Proceedings 1995, pp.  331–339. Morgan Kaufmann, San Francisco (CA), 1995. ISBN 978-1-55860-377-6. doi: https://doi.org/10.1016/B978-1-55860-377-6.50048-7. URL https://www.sciencedirect.com/science/article/pii/B9781558603776500487.
  41. A comprehensive survey to dataset distillation. arXiv preprint arXiv:2301.05603, 2023.
  42. Evaluating Modern GPU Interconnect: PCIe, NVLink, NV-SLI, NVSwitch and GPUDirect. IEEE Transactions on Parallel and Distributed Systems, 31(1):94–110, January 2020a. ISSN 2161-9883. doi: 10.1109/tpds.2019.2928289. URL http://dx.doi.org/10.1109/TPDS.2019.2928289.
  43. Federated Optimization in Heterogeneous Networks. In Dhillon, I., Papailiopoulos, D., and Sze, V. (eds.), Proceedings of Machine Learning and Systems, volume 2, pp.  429–450, 2020b. URL https://proceedings.mlsys.org/paper_files/paper/2020/file/1f5fe83998a09396ebe6477d9475ba0c-Paper.pdf.
  44. Fair Resource Allocation in Federated Learning. In International Conference on Learning Representations, 2020c. URL https://openreview.net/forum?id=ByexElSYDr.
  45. FedBN: Federated Learning on Non-IID Features via Local Batch Normalization. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=6YEQUn0QICG.
  46. Privacy-preserving traffic flow prediction: A federated learning approach. IEEE Internet of Things Journal, 7(8):7751–7763, 2020.
  47. Learning under Concept Drift: A Review. IEEE Transactions on Knowledge and Data Engineering, 31(12):2346–2363, 2019. doi: 10.1109/TKDE.2018.2876857.
  48. Communication-Efficient Learning of Deep Networks from Decentralized Data. In Singh, A. and Zhu, J. (eds.), Proceedings of the 20th International Conference on Artificial Intelligence and Statistics, volume 54 of Proceedings of Machine Learning Research, pp.  1273–1282. PMLR, 20–22 Apr 2017a. URL https://proceedings.mlr.press/v54/mcmahan17a.html.
  49. Learning Differentially Private Recurrent Language Models, 2017b. URL https://arxiv.org/abs/1710.06963.
  50. Cefl: Carbon-efficient federated learning, 2023. URL https://arxiv.org/abs/2310.17972.
  51. A survey on security and privacy of federated learning. Future Generation Computer Systems, 115:619–640, 2021.
  52. Nature. There are holes in Europe’s AI Act — and researchers can help to fill them. Nature, 625(7994):216–216, January 2024. ISSN 1476-4687. doi: 10.1038/d41586-024-00029-4. URL http://dx.doi.org/10.1038/d41586-024-00029-4.
  53. Federated learning in a medical context: a systematic literature review. ACM Transactions on Internet Technology (TOIT), 21(2):1–31, 2021.
  54. Prechelt, L. Early stopping-but when? In Neural Networks: Tricks of the trade, pp.  55–69. Springer, 2002.
  55. A first look into the carbon footprint of federated learning. Journal of Machine Learning Research, 24(129):1–23, 2023. URL http://jmlr.org/papers/v24/21-0445.html.
  56. Adaptive Federated Optimization, 2020. URL https://arxiv.org/abs/2003.00295.
  57. Energy Mix. Our World in Data, 2020. https://ourworldindata.org/energy-mix.
  58. Trusted Execution Environment: What It is, and What It is Not. In 2015 IEEE Trustcom/BigDataSE/ISPA. IEEE, August 2015. doi: 10.1109/trustcom.2015.357. URL http://dx.doi.org/10.1109/Trustcom.2015.357.
  59. Energy-Efficient Federated Learning With Resource Allocation for Green IoT Edge Intelligence in B5G. IEEE Access, 11:16353–16367, 2023. doi: 10.1109/ACCESS.2023.3244099.
  60. Laion-5b: An open large-scale dataset for training next generation image-text models. Advances in Neural Information Processing Systems, 35:25278–25294, 2022.
  61. Towards Personalized Federated Learning. IEEE Transactions on Neural Networks and Learning Systems, 34(12):9587–9603, December 2023. ISSN 2162-2388. doi: 10.1109/tnnls.2022.3160699. URL http://dx.doi.org/10.1109/TNNLS.2022.3160699.
  62. The European Commission. Cookie Pledge, 2023. URL https://commission.europa.eu/live-work-travel-eu/consumer-rights-and-complaints/enforcement-consumer-protection/cookie-pledge_en.
  63. The White House. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, 10 2023. URL https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/.
  64. FedBERT: When Federated Learning Meets Pre-training. ACM Transactions on Intelligent Systems and Technology, 13(4):1–26, August 2022. ISSN 2157-6912. doi: 10.1145/3510033. URL http://dx.doi.org/10.1145/3510033.
  65. Unbiased look at dataset bias. In CVPR 2011, pp.  1521–1528. IEEE, 2011.
  66. Federated learning based energy demand prediction with clustered aggregation. In 2021 IEEE International Conference on Big Data and Smart Computing (BigComp), pp.  164–167. IEEE, 2021.
  67. Energy Consumption Comparison of Interactive Cloud-Based and Local Applications. IEEE Journal on Selected Areas in Communications, 33(4):616–626, April 2015. ISSN 0733-8716. doi: 10.1109/jsac.2015.2393431. URL http://dx.doi.org/10.1109/JSAC.2015.2393431.
  68. Electricity consumer characteristics identification: A federated learning approach. IEEE Transactions on Smart Grid, 12(4):3637–3647, 2021.
  69. Data collection and quality challenges in deep learning: A data-centric ai perspective. The VLDB Journal, 32(4):791–813, 2023.
  70. Fedzero: Leveraging renewable excess energy in federated learning. 2023. doi: 10.48550/ARXIV.2305.15092. URL https://arxiv.org/abs/2305.15092.
  71. FLEdge: Benchmarking Federated Machine Learning Applications in Edge Computing Systems. arXiv preprint arXiv:2306.05172, 2023.
  72. Machine Unlearning: A Survey. ACM Comput. Surv., 56(1), aug 2023. ISSN 0360-0300. doi: 10.1145/3603620. URL https://doi.org/10.1145/3603620.
  73. Client Selection and Bandwidth Allocation in Wireless Federated Learning Networks: A Long-Term Perspective. IEEE Transactions on Wireless Communications, 20(2):1188–1200, February 2021. ISSN 1558-2248. doi: 10.1109/twc.2020.3031503. URL http://dx.doi.org/10.1109/TWC.2020.3031503.
  74. Green Federated Learning, 2023. URL https://arxiv.org/abs/2303.14604.
  75. A survey on federated learning. Knowledge-Based Systems, 216:106775, March 2021. ISSN 0950-7051. doi: 10.1016/j.knosys.2021.106775. URL http://dx.doi.org/10.1016/j.knosys.2021.106775.
  76. Efficient Membership Inference Attacks against Federated Learning via Bias Differences. In Proceedings of the 26th International Symposium on Research in Attacks, Intrusions and Defenses, RAID 2023. ACM, October 2023. doi: 10.1145/3607199.3607204. URL http://dx.doi.org/10.1145/3607199.3607204.
  77. FedPrompt: Communication-Efficient and Privacy-Preserving Prompt Tuning in Federated Learning. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp.  1–5. IEEE, 2023.
  78. Federated learning with non-iid data. arXiv preprint arXiv:1806.00582, 2018.
Citations (8)
View on Semantic Scholar

Summary

  • The paper demonstrates that Federated Learning can align with stringent EU AI regulations by enhancing data governance and mitigating bias without compromising privacy.
  • It reveals that energy efficiency challenges require innovative optimization to balance privacy safeguards with computational overhead.
  • The study prioritizes robust model validation and compliance frameworks, guiding future research for scalable and legally compliant AI systems.

Federated Learning Priorities Under the European Union Artificial Intelligence Act

The paper "Federated Learning Priorities Under the European Union Artificial Intelligence Act" provides an intricate analysis of how the European Union's Artificial Intelligence Act could impact the field of Federated Learning (FL). The discourse reveals the necessity for the FL community to reconsider its research goals to align with the demands of this legal framework, thereby positioning itself as a key player in developing AI systems compliant with privacy and data governance standards.

The manuscript begins by contextualizing the emergence of the AI Act, established by the EU Commission and Parliament, which is designed to impose rigorous requirements on machine learning applications. Federated Learning is presented as a method that contrasts sharply with centralized machine learning approaches due to its inherent focus on data privacy across decentralized networks. Unlike traditional methods, FL does not require raw data to leave the client’s device, yet this process poses unique challenges and opportunities under the emerging regulations.

The document systematically breaks down the core implications of the AI Act for FL into several critical areas:

  1. Data Governance: The AI Act emphasizes stringent oversight of data bias and mandates compliance with GDPR. Federated Learning, inherently designed to preserve privacy, already aligns with aspects of these requirements. However, challenges such as addressing bias without directly accessing data must be tackled. The authors highlight the potential of FL to expand data access without compromising privacy, a key compliance factor.
  2. Energy Efficiency: The act anticipates new standards for energy-efficient AI systems. Federated Learning, which uses less powerful edge devices in contrast to data center GPUs, enters a nuanced landscape where its energy usage and sustainability need to be quantified. This is further complicated by the privacy-energy trade-offs introduced by adopting private computation techniques, which can increase computational overhead.
  3. Robustness and Quality Management: To meet the high-risk application criteria under the AI Act, FL systems need to maintain high accuracy and robustness. The text implies a need for innovative approaches to validate models frequently without incurring significant energy and communication costs, such as asynchronous validation techniques.

The paper also provides an empirical and qualitative examination of the FL model prototypes that adhere to the AI Act. The experimental results illustrate the trade-offs between computational efficiency and privacy guarantees when adopting FL systems compliant with the legal framework.

In anticipation of practical deployment, the authors propose several research priorities as foundational steps for the federated learning community. These include focusing on optimizing data quality and bias detection methods, enhancing energy efficiency through novel optimization strategies, and developing comprehensive frameworks that provide regulatory compliance guidelines.

The authors postulate that FL has the potential to serve as a preferred method for ML systems under the AI Act due to its privacy-preserving nature and ability to handle non-IID data. Nevertheless, challenges remain, including the need to align arithmetic privacy with legal norms and to address the economic cost imbalances between FL and centralized systems.

Overall, the paper provides a thorough examination of the ramifications of the AI Act on Federated Learning, outlining existing hurdles while inspiring future research aimed at making FL a compliant and dominant methodology for privacy-preserving AI applications. The authors make compelling arguments for the FL community to engage vigorously with the regulatory landscape to ensure that the benefits of the technology align with societal values as expressed in the legislation.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Sign Up to Generate All Videos Subscribe on YouTube

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Sign Up to Generate

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Generate Now

Continue Learning

We haven't generated follow-up questions for this paper yet.

Generate Now

Collections

Sign up for free to add this paper to one or more collections.

Sign Up

Tweets

Sign up for free to view the 1 tweet with 0 likes about this paper.

Sign Up for Free