Privacy and Data Control on Social Networks using Deep Learning
A. Parimala, Y. Vijayalata · 2022
Online social networks are huge data exchange platforms that help to promote and share a lot of good information about products, news, education, tourism, health care, etc., also there is a great risk involved to individual’s privacy and security. Online posted photos can be shared, tagged, and reposted again without having any consent from the people present in the images. In this project, we are proposing a facial recognition-based system that detects every single individual and checks their online account relationship between the individual. The post or image that is reposted, tagged, or copied by known or unknown individuals will be checked. If it is an unknown individual the images convolved through a low-pass filter kernel which removes the high-frequency content like noise and edges to smooth the images that make the image blur. For this mechanism, we proposed to use deep learning based VGG19 architecture for image classification and Haar cascade viola jones algorithm for detection, and computer vision-based blurring techniques used. algorithm performance as the area under the ROC curve (AUC) accuracy of 88% obtained. This proposed model achieved 91.53% classification accuracy and further we evaluated the model using Accuracy, Precision, Recall and F1-Score metrices and achieved 70.83%, 82%, 71% and 0.68%.