Effectiveness in Collaborative Framework for Non-Invasive in AI Algorithms
Sandeep Kulkarni1, B.Vijayendra Reddy2
1Dr. Sandeep Kulkarni, Software Developer, Pune (Maharashtra), India.
2B.Vijayendra Reddy, Department of Computer Science Engineering, Lovely Professional University, Phagwara (Punjab), India.
Manuscript received on 01 March 2024 | Revised Manuscript received on 10 March 2024 | Manuscript Accepted on 15 March 2024 | Manuscript published on 30 March 2024 | PP: 16-19 | Volume-14 Issue-1, March 2024 | Retrieval Number: 100.1/ijsce.F451713060824 | DOI: 10.35940/ijsce.F4517.14010324
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The topic of study and practice known as “privacy-preserving machine learning (PPML)” is devoted to creating methods and strategies that enable the training and application of machine learning models while protecting the privacy of sensitive data for convolution neural network and Machine learning algorithms. Garbled worlds” is a concept primarily used in the context of privacy-preserving machine learning (PPML). It refers to a technique used to protect the privacy of individual data points during the training process of machine learning models. Garbled worlds allow organizations or individuals to collaborate and train machine learning models using their combined datasets without sharing the raw data. This is particularly important in scenarios where data privacy regulations or concerns prohibit the sharing of sensitive information. By using garbled worlds, organizations can leverage the collective knowledge in multiple datasets while protecting the privacy of individuals whose data contributes to the training process. This technique helps balance data privacy and the utility of machine learning models in various applications. The effectiveness and adaptability of ABY3 (The mixed protocol framework for machine learning) enable users to select several cryptographic protocols based on their unique needs and limitations. In comparison to other safe multi-party computation frameworks, it minimizes computational and communication costs while maintaining a high level of security. The viability of our system is demonstrated by the enhanced benchmarking of the previously described algorithms in contrast to ABY3.
Keywords: ABY3, MLaaS, GDPR, Homomorphic Encryption, Logistic Regression, Linear Regression, Convolution Neural Network.
Scope of the Article: Neural Networking