Browse Prior Art Database

Intelligent Machine Learning Algorithm to Protect Privacy

IP.com Disclosure Number: IPCOM000250567D
Publication Date: 2017-Aug-03
Document File: 3 page(s) / 41K

Publishing Venue

The IP.com Prior Art Database

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 50% of the total text.

Intelligent Machine Learning Algorithm to Protect Privacy

Abstract

An intelligent mechanism where user can select few/complete parts of the image/video to prevent any machine learning algorithm to use such content for training or classification. The machine learning algorithm while processing the content will look for the parts which are selected by the user and skips them. This can be used by social networking sites, storage companies, cloud based companies which store the images or videos uploaded by the user.

Introduction

Artificial Intelligence or Cognitive Computing is the latest trend where a system is learned with lot of data and later any input to the system is classified to give an output based on the learning. Consider Watson's Visual Recognition's API which lets you know what is present in the image. For example, if a system is trained with all the images about cricket. Any image input to this API will recognize all the cricket related objects in the image. Features like this will also help in image searching. If an image is provided as input to a search algorithm, it can easily search similar or related images. Such intelligent features may lead to confidentiality breach or privacy invasion if not properly utilized. Everyday user's upload multiple photos to the social networking sites or blogs. For example, Facebook gets 350 million photos uploaded every day. These photos can be used by the respective site for various purposes. A machine learning algorithm at a site database server can process these images and store all the contents of the image along with the data. When user searches for a content, it can display all images with such content. In the process, no filtering happens and confidentiality or privacy breaches may occur as there is no existing mechanism for the user to convey that a certain image or certain parts of the image cannot be given to a machine learning system. We propose an intelligent way for the user to convey to machine learning algorithm that a certain content should not be used for training or classification. An intelligent mechanism is proposed where user can select few/complete parts of the image/video to prevent any machine learning algorithm to use such content for training or classification. The machine learning algorithm while processing the content will look for the parts which are selected by the user and skips them. This can be used by social networking sites, storage companies, cloud based companies which store the images or videos uploaded by the user.

Working of the proposed method:

The detailed description works in following steps: 1. Any site or service provider will provide an interface to upload the image. Before the image is actually uploaded to server, an interface is displayed to the user to select parts of the image that should not be considered for training or classification. For example, say user is uploading a photo of his near some military ammunition base, he can select the parts o...