Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{186115,
        author = {T P SNEHA},
        title = {Homomorphic Encryption for Deep Neural Networks: A Privacy-Preserving Paradigm for Secure AI Computation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {6},
        pages = {4506-4508},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=186115},
        abstract = {As Artificial Intelligence (AI) systems increasingly operate on sensitive data—such as medical records, financial transactions, and biometric identifiers—data privacy has become a central concern. Homomorphic Encryption (HE) offers a cryptographic mechanism that enables computation directly on encrypted data without decryption, preserving privacy while maintaining utility. This paper presents a comprehensive study and implementation framework of integrating Homomorphic Encryption with Deep Neural Networks (HE-DNN). We explore efficient encryption schemes compatible with matrix operations, propose an optimized HE-friendly neural network architecture, and introduce a computational optimization strategy that reduces latency and ciphertext expansion. Experimental results on benchmark datasets demonstrate the feasibility of training and inference on encrypted data with minimal accuracy degradation. The proposed framework advances privacy-preserving AI, enabling secure cloud-based machine learning services without exposing raw data.},
        keywords = {Homomorphic Encryption, Deep Neural Networks, Privacy-Preserving AI, Secure Computation, Federated Learning, Cryptographic AI, Encrypted Inference},
        month = {November},
        }