This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
Raina Singh,
Gungun Verma,
Arpit Kumar,
Arun Kumar Rawat,
Hardik Sharma,
- Student, Computer science and engineering, Greater Noida Institute of Technology, Greater Noida, Uttar Pradesh, India
- Student, Computer science and engineering, Greater Noida Institute of Technology, Greater Noida, Uttar Pradesh, India
- Student, Computer science and engineering, Greater Noida Institute of Technology, Greater Noida, Uttar Pradesh, India
- Student, Computer science and engineering, Greater Noida Institute of Technology, Greater Noida, Uttar Pradesh, India
- Student, Computer science and engineering, Greater Noida Institute of Technology, Greater Noida, Uttar Pradesh, India
Abstract
Okay so here is a problem that has been bothering us for a while. Everyone talks about how data is the new oil, how data-driven decisions are transforming healthcare, finance, governance, all of it. And that is true. But there is this massive elephant in the room that a lot of people in the data science community sort of wave away: privacy. The more useful the data, the more personal it tends to be. And the techniques people have relied on for years to “anonymize” data, things like removing names or masking IDs, have been shown to fail repeatedly against modern re- identification attacks. Regulations have caught up. The GDPR in Europe, India’s Digital Personal Data Protection Act, and similar laws around the world are forcing organizations to actually take privacy seriously rather than treating it as an afterthought. But compliance is hard when your existing data pipelines were never designed with privacy in mind. This is where Privacy-by-Design comes in — the idea that privacy should be baked into the system architecture from the start, not bolted on later. In this paper, we review two technologies that we think represent the most promising directions for making Privacy-by-Design actually work in practice: Differential Privacy, which adds carefully Page 2 of 2 calibrated noise to data or query outputs so individual records cannot be reverse-engineered, and Federated Analytics, which lets you compute insights from data that never leaves the device it was collected on. We go through how each of these works, where they have been deployed in the real world, what their limitations are, and how they compare against traditional anonymization approaches. We also look at where these two ideas intersect and whether combining them could give us something better than either one alone.
Keywords: Privacy-by-Design, Differential Privacy, Federated Analytics, Federated Learning, Data Protection, GDPR, DPDP Act, k-Anonymity, Data Science, Privacy-Preserving Machine Learning
References
[1] Narayanan A, Shmatikov V. Robust De-anonymization of Large Sparse Datasets. 2008 IEEE Symposium on Security and Privacy (sp 2008). 2008 May;111–25. Available from: https://ieeexplore.ieee.org/document/4531148 .
[2] Sweeney L. k-anonymity: A model for protecting privacy. International journal of uncertainty, fuzziness and knowledge-based systems. 2002 Oct;10(05):557-70.
[3] Protection D. General data protection regulation. Intersoft Consulting, Accessed in October. 2018 May 25;24(1).
[4] The Digital Personal Data Protection Bill, 2023 [Internet]. PRS Legislative Research. 2026 [cited 2026 Aug 31]. Available from: https://prsindia.org/billtrack/digital-personal-data-protection-bill-2023.
[5] Dwork C, McSherry F, Nissim K, Smith A. Calibrating noise to sensitivity in private data analysis. InTheory of cryptography conference 2006 Mar 4 (pp. 265-284). Berlin, Heidelberg: Springer Berlin Heidelberg.
[6] McMahan B, Moore E, Ramage D, Hampson S, y Arcas BA. Communication-efficient learning of deep networks from decentralized data. InArtificial intelligence and statistics 2017 Apr 10 (pp. 1273-1282). Pmlr.
[7] Machanavajjhala A, Kifer D, Gehrke J, Venkitasubramaniam M. l-diversity: Privacy beyond k-anonymity. Acm transactions on knowledge discovery from data (tkdd). 2007 Mar 1;1(1):3-es.
[8] Abowd JM. The US Census Bureau adopts differential privacy. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining 2018 Jul 19 (pp. 2867-2867). Page 9 of 2
[9] Mironov I. Rényi differential privacy. In2017 IEEE 30th computer security foundations symposium (CSF) 2017 Aug 21 (pp. 263-275). IEEE.
[10] Erlingsson Ú, Pihur V, Korolova A. Rappor: Randomized aggregatable privacy-preserving ordinal response. InProceedings of the 2014 ACM SIGSAC conference on computer and communications security 2014 Nov 3 (pp. 1054-1067).
[11] Federated Analytics: Collaborative Data Science without Data Collection. Research.google. 2020. Available from: https://research.google/blog/federated-analytics-collaborative-data-science-without-data-collection/.
[12] Bonawitz K, Ivanov V, Kreuter B, Marcedone A, McMahan HB, Patel S, Ramage D, Segal A, Seth K. Practical secure aggregation for privacy-preserving machine learning. Inproceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security 2017 Oct 30 (pp. 1175-1191).
[13] Melis L, Song C, De Cristofaro E, Shmatikov V. Exploiting unintended feature leakage in collaborative learning. In2019 IEEE symposium on security and privacy (SP) 2019 May 19 (pp. 691-706). IEEE.
[14] Lee J, Clifton C. How much is enough? choosing ε for differential privacy. InInternational Conference on Information Security 2011 Oct 26 (pp. 325-340). Berlin, Heidelberg: Springer Berlin Heidelberg.
| Volume | 04 | |
| 02 | ||
| Received | 08/07/2026 | |
| Accepted | 27/07/2026 | |
| Published | 07/09/2026 | |
| Publication Time | 61 Days |