IRO Journals
Journal of ISMAC

LSAB-zCDP: Layer-wise Sensitivity Analysis with Tight Bounds under Zero-Concentrated Differential Privacy for Deep Autoencoder Composition in Medical IoT

Open Access
Published: 29 June, 2026
Pages: 224-248
  • Manas Kumar Yogi , Manas Kumar Yogi
    Department of Computer Science and Engineering, JNTUK College of Engineering, Kakinada, Andhra Pradesh, India
    Department of Computer Science and Engineering, JNTUK College of Engineering, Kakinada, Andhra Pradesh, India
  • Chakravarthy A.S.N. Chakravarthy A.S.N.
    Department of Computer Science and Engineering, JNTUK College of Engineering, Kakinada, Andhra Pradesh, India
    Department of Computer Science and Engineering, JNTUK College of Engineering, Kakinada, Andhra Pradesh, India
view PDF
view PDF

How to Cite

Yogi, Manas Kumar, and Chakravarthy A.S.N. 2026. “LSAB-ZCDP: Layer-Wise Sensitivity Analysis With Tight Bounds under Zero-Concentrated Differential Privacy for Deep Autoencoder Composition in Medical IoT”. Journal of ISMAC 8 (3): 224-48. https://doi.org/10.36548/jismac.2026.3.003.
ACM ACS APA ABNT Chicago Harvard IEEE MLA Turabian Vancouver AMA

Download Citation

Endnote/Zotero/Mendeley (RIS) BibTeX

Keywords

Zero-Concentrated Differential Privacy
zCDP
Deep Autoencoder
Layer-wise Sensitivity
Lipschitz Bound
ReLU
GELU
Medical IoT
PTB-XL
Physiological Signal Processing

Abstract

Deep autoencoders comprising stacked convolutional, residual, and transformer layers are state-of-the-art architectures for physiological signal representation in Medical Internet of Things (MIoT) systems. Applying zero-Concentrated Differential Privacy (zCDP) to such deep architectures is fundamentally challenging due to the need for accurate layer-wise sensitivity bounds; existing global Lipschitz-based bounds are provably loose, resulting in over-injection of Gaussian noise that degrades clinical utility. This research presents LSAB-zCDP, a Layer-wise Sensitivity Analysis with tight Bounds framework under zCDP for deep autoencoder composition in MIoT. The framework proceeds in four phases: (1) Distributional Sensitivity Profiling estimating empirical pre-activation distributions and computing activation-specific local Lipschitz constants; (2) Tight Bound Computation and Pareto-optimal Noise Allocation recursive computation of tight per-layer bounds and Lagrangian-derived noise scales; (3) DP-SGD Training with Layer-Adaptive Clipping replacing single global noise with layer-specific calibration; and (4) Privacy Certificate Issuance formal (ε, δ)-DP conversion for HIPAA reporting. Evaluated on PTB-XL, OhioT1DM, and MIMIC-III Waveform datasets, LSAB-zCDP achieves AUC-ROC of 0.891 at ε = 1.0 outperforming the strongest baseline by 17.1% while consuming 28.6% less privacy budget than loose-bound methods at identical utility.

References

  1. Abadi, Martin, Andy Chu, Ian Goodfellow, H. Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. "Deep Learning with Differential Privacy." In Proceedings of the 2016 ACM SIGSAC conference on computer and communications security 2016: 308-318.
  2. Anil, Rohan, Badih Ghazi, Vineet Gupta, Ravi Kumar, and Pasin Manurangsi. "Large-Scale Differentially Private BERT." In Findings of the Association for Computational Linguistics: EMNLP 2022: 6481-6491.
  3. Beaulieu-Jones, Brett K., Zhiwei Steven Wu, Chris Williams, Ran Lee, Sanjeev P. Bhavnani, James Brian Byrd, and Casey S. Greene. "Privacy-Preserving Generative Deep Neural Networks Support Clinical Data Sharing." Circulation: Cardiovascular Quality and Outcomes 2019, vol 12, no. 7: e005122.
  4. Bun, Mark, Cynthia Dwork, Guy N. Rothblum, and Thomas Steinke. "Composable and Versatile Privacy via Truncated Cdp." In Proceedings of the 50th Annual ACM SIGACT Symposium on Theory of Computing 2018: 74-86.
  5. Bun, Mark, and Thomas Steinke. "Concentrated Differential Privacy: Simplifications, Extensions, and Lower Bounds." In Theory of cryptography conference, Berlin, Heidelberg: Springer Berlin Heidelberg, 2016: 635-658.
  6. Dong, Jinshuo, Aaron Roth, and Weijie J. Su. "Gaussian Differential Privacy." Journal of the Royal Statistical Society Series B: Statistical Methodology 2022, vol 84, no. 1: 3-37.
  7. Dwork, Cynthia, and Aaron Roth. "The Algorithmic Foundations of Differential Privacy." Foundations and trends® in theoretical computer science 2014, vol 9, no. 3-4: 211-487.
  8. Dwork, Cynthia, and Guy N. Rothblum. "Concentrated Differential Privacy." arXiv preprint 2016, arXiv:1603.01887.
  9. Goodfellow, Ian, Yoshua Bengio, Aaron Courville, and Yoshua Bengio. Deep Learning. Vol. 1, no. 2. Cambridge: MIT press, 2016: 351-354.
  10. He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep Residual Learning for Image Recognition." In Proceedings of the IEEE conference on computer vision and pattern recognition 2016: 770-778.
  11. Hendrycks, Dan, and Kevin Gimpel. "Gaussian Error Linear Units (gelus)." arXiv preprint 2016, arXiv:1606.08415.
  12. Hoory, Shlomo, Amir Feder, Avichai Tendler, Sofia Erell, Alon Peled-Cohen, Itay Laish, Hootan Nakhost et al. "Learning and Evaluating a Differentially Private Pre-Trained Language Model." In Findings of the Association for Computational Linguistics: EMNLP 2021: 1178-1189.
  13. Ioffe, Sergey, and Christian Szegedy. "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift." In International conference on machine learning, Proceedings of Machine Learning Research 2015: 448-456.
  14. Johnson, Alistair EW, Tom J. Pollard, Lu Shen, Li-wei H. Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo Anthony Celi, and Roger G. Mark. "MIMIC-III, a Freely Accessible Critical Care Database." Scientific data 2016, vol 3, no. 1: 1-9.
  15. Klambauer, Günter, Thomas Unterthiner, Andreas Mayr, and Sepp Hochreiter. "Self-Normalizing Neural Networks." Advances in neural information processing systems 2017, vol 30: 971–980.
  16. Lecuyer, Mathias, Vaggelis Atlidakis, Roxana Geambasu, Daniel Hsu, and Suman Jana. "Certified Robustness to Adversarial Examples with Differential Privacy." arXiv preprint 2018, arXiv:1802.03471: 656–672.
  17. Marling, Cindy, and Razvan Bunescu. "The OhioT1DM Dataset for Blood Glucose Level Prediction: Update 2020." In CEUR workshop proceedings 2020, vol. 2675: 71-85
  18. McMahan, H. Brendan, Daniel Ramage, Kunal Talwar, and Li Zhang. "Learning Differentially Private Recurrent Language Models." arXiv preprint 2017, arXiv:1710.06963.
  19. Mironov, Ilya. "Rényi Differential Privacy." In 2017 IEEE 30th computer security foundations symposium (CSF), IEEE, 2017: 263-275.
  20. Tsuzuku, Yusuke, Issei Sato, and Masashi Sugiyama. "Lipschitz-Margin Training: Scalable Certification of Perturbation Invariance for Deep Neural Networks." Advances in neural information processing systems 2018, vol 31: 6541–6550.
  21. Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. "Attention is All You Need." Advances in neural information processing systems 2017, vol 30: 5998–6008.
  22. Virmaux, Aladin, and Kevin Scaman. "Lipschitz Regularity of Deep Neural Networks: Analysis and Efficient Estimation." Advances in neural information processing systems 2018, vol 31: 3835–3844.
  23. Wagner, Patrick, Nils Strodthoff, Ralf-Dieter Bousseljot, Dieter Kreiseler, Fatima I. Lunze, Wojciech Samek, and Tobias Schaeffter. "PTB-XL, a Large Publicly Available Electrocardiography Dataset." Scientific data 2020, vol 7, no. 1: 154.
  24. Wang, Yu-Xiang, Borja Balle, and Shiva Prasad Kasiviswanathan. "Subsampled Rényi Differential Privacy and Analytical Moments Accountant." In The 22nd international conference on artificial intelligence and statistics, Proceedings of Machine Learning Research 2019: 1226-1235.
  25. Stoian, Mihaela CÄ, Salijona Dyrmishi, Maxime Cordy, Thomas Lukasiewicz, and Eleonora Giunchiglia. "How Realistic is Your Synthetic Data? Constraining Deep Generative Models for Tabular Data." arXiv preprint 2024, arXiv:2402.04823.