An Optimization Study of TFHE Encryption Algorithm Based on tiny-DNN
Main Article Content
Keywords
deep learning framework, fully homomorphic encryption, data security, MUX logic gate
Abstract
Privacy leakage, malicious tampering, and incompleteness of big data have become severe challenges for deep learning. Meanwhile, fully homomorphic encryption algorithms suffer from increased energy consumption and insufficient precision. To address these issues, this paper proposes a dual-path heterogeneous privacy-preserving inference mechanism. First, data is divided into private and non-private categories using a privacy segmentation coefficient. Private data is placed in the TFHE encryption domain, while a tiny-DNN-based secure black-box is employed to protect non-private data. The zero-error nonlinear activation mechanism of the MUX logic gate in TFHE is used to train the deep learning framework. Subsequently, a cross-domain fixed-point homomorphic adder fuses the two types of data to obtain the ciphertext result formed by end-to-end heterogeneous data. This approach ensures the prediction accuracy of deep learning while reducing the network latency of big data classification. Experimental results show that the proposed system reduces end-to-end inference latency to 970 ms while maintaining model accuracy. Security analysis demonstrates that the system possesses strong robustness against attacks, thereby providing confidentiality and integrity protection for real-time inference of big data.
References
- [1] Yang Li, Ruinong Wang, Yuanzheng Li, Meng Zhang, and Chao Long. Wind power forecasting considering data privacy protection: A federated deep reinforcement learning approach. Applied Energy, 329:120291, 2023.
- [2] Sara Quach, Park Thaichon, Kelly D Martin, Scott Weaven, and Robert W Palmatier. Digital technologies: tensions in privacy and data. Journal of the academy of marketing science, 50(6):1299–1323, 2022.
- [3] Ronald L Rivest, Len Adleman, Michael L Dertouzos, et al. On data banks and privacy homomorphisms. Foundations of secure computation, 4(11):169–180, 1978.
- [4] Kundan Munjal and Rekha Bhatia. A systematic review of homomorphic encryption and its contributions in healthcare industry. Complex & Intelligent Systems, 9(4):3759–3786, 2023.
- [5] Alexander Viand, Christian Knabenhans, and Anwar Hithnawi. Verifiable fully homomorphic encryption. arXiv preprint arXiv:2301.07041, 2023.
- [6] Wei Ao and Vishnu Naresh Boddeti. AutoFHE: Automated adaption of CNNs for efficient evaluation over FHE. In 33rd USENIX Security Symposium (USENIX Security 24), pages 2173–2190, 2024.
- [7] Ershat Arkin, Nurbiya Yadikar, Xuebin Xu, Alimjan Aysa, and Kurban Ubul. A survey: object detection methods from cnn to transformer. Multimedia Tools and Applications, 82(14):21353–21383, 2023.
- [8] Lars Folkerts, Charles Gouert, and Nektarios Georgios Tsoutsos. REDsec: Running encrypted discretized neural networks in seconds. Cryptology ePrint Archive, 2021.
- [9] Abdulrahman Diaa, Lucas Fenaux, Thomas Humphries, Marian Dietz, Faezeh Ebrahimianghazani, Bailey Kacsmar, Xinda Li, Nils Lukas, Rasoul Akhavan Mahdavi, Simon Oya, et al. Fast and private inference of deep neural networks by co-designing activation functions. In 33rd USENIX Security Symposium (USENIX Security 24), pages 2191–2208, 2024.
- [10] Kai Shen, Junliang Guo, Xu Tan, Siliang Tang, Rui Wang, and Jiang Bian. A study on relu and softmax in transformer. arXiv preprint arXiv:2302.06461, 2023.
- [11] Tim Martin and Frank Allgöwer. Data-driven system analysis of nonlinear systems using polynomial approximation. IEEE Transactions on Automatic Control, 69(7):4261–4274, 2023.
- [12] Xiaowei Yu, Zhe Huang, Yao Xue, Lu Zhang, Li Wang, Tianming Liu, and Dajiang Zhu. Noisynn: Exploring the influence of information entropy change in learning systems. arXiv preprint arXiv:2309.10625, 2023.
- [13] Umberto Garlando, Qi Wang, Oleksandr V Dobrovolskiy, Andrii V Chumak, and Fabrizio Riente. Numerical model for 32-bit magnonic ripple carry adder. IEEE Transactions on Emerging Topics in Computing, 11(3):679–688, 2023.
- [14] Muhanna Ahmed Ali and Salah Alawi Hussein Al-Sharafi. Intrusion detection in iot networks using machine learning and deep learning approaches for mitm attack mitigation. Discover Internet of Things, 5(1):48, 2025.
- [15] Dmytro Tymoshchuk, Oleh Yasniy, Mykola Mytnyk, Nataliya Zagorodna, and Vitaliy Tymoshchuk. Detection and classification of ddos flooding attacks by machine learning method. arXiv preprint arXiv:2412.18990, 2024.
- [16] S Abiramasundari and V Ramaswamy. Distributed denial-of-service (ddos) attack detection using supervised machine learning algorithms. Scientific Reports, 15(1):13098, 2025.
Article Sidebar
This work is licensed under a Creative Commons Attribution 4.0 International License.