Privacy Preservation using Deep Learning Methods
Amreen Ayisha M1, Ankit Kumar2, Gasti Harshini3, Keerthana S4, Mrs. Samatha R Swamy5
12345 ayishaamreen78@gmail.com, ankit.astaim@gmail.com, harshinisweety61@gmail.com, twinklekee026@gmail.com, samatha-ise@dsatm.edu.in
1234Student, Department of Information Science and Engineering, DSATM, Bangalore-88, Karnataka
5Faculty, Department of Information Science and Engineering, DSATM, Bangalore-88, Karnataka
Abstract- Significant technological advancements have occurred over the past 10 years, which have the ability to make daily living routines more convenient on both a corporate and personal level. The sharing of data with other global web applications that keep tabs on your daily actions, however, is the next real-life problem. Building a collaborative platform and maintaining individual privacy are the main concerns for training a global Deep Learning (DL) model. Deep learning is a subset of machine learning that makes use of multiple-layered neural networks to extract patterns from data. This technology has been used to develop several privacy-preserving techniques, including data anonymization, differential privacy, and homomorphic encryption. Deep Neural Network (DNN) has been appearing incredible potential in sorts of real-world applications such as extortion discovery and trouble expectation. In the meantime, information separation has gotten to be a genuine issue right now, i.e., diverse parties cannot share information with each other. To unravel this issue, most investigate leverages cryptographic strategies to prepare secure DNN models for multi-parties without compromising their private information. In spite of the fact that such strategies have solid security ensure, they are troublesome to scale to profound systems and expansive datasets due to its tall communication and computation complexities.
Keywords- Differential Privacy, Adversarial Learning, Secret Sharing, Encrypted Deep Learning, Security, Federated Learning.
