Browse Prior Art Database

a apparatus and method to detect pictures with sensitive information based on mobile phone picture context and features

IP.com Disclosure Number: IPCOM000244603D
Publication Date: 2015-Dec-28
Document File: 6 page(s) / 143K

Publishing Venue

The IP.com Prior Art Database

Abstract

Mobile phone has become the most commonly used device to take photos in life. For convenience, people would like to take photos with some personal sensitive information (such identity card, credit card, passport and so on) as a memorandum so that the sensitive information is very easy to be exposed through internet. We proposed a method to detect the sensitive information for mobile photos and trigger alert before sending to internet. Compared with traditional image sensitive detection, we focus on not only the features from the image itself but also the image ’s context on mobile phone such as focus length or neighbor images’ sensitivity. For trigger alert part, we provided an automatic confidence-based tagging and altering decider to prevent sensitive image leakage.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 43% of the total text.

Page 01 of 6

a apparatus and method to detect pictures with sensitive information based on mobile phone picture context and features

Background:

Mobile phone has become the most commonly used device to take photos in people's life. People use mobile phone to take photos everywhere, even including Identity Card, Credit Card and etc. So security on mobile phonebecomes very important and an automatic altering and different protection policy for images is a necessary to prevent sensitive information leakage.

Problem:

1. Personal sensitive information is very easy to be exposed through internet
-People use mobile phone to take photos that contain sensitive information
-These kinds of information will not be identified and then will be transferred to internet without any alerting

2. Imaged-based identification algorithm is complex
-Resource on mobile on is limited on mobile phone
-For the sake of power saving, complex traditional Image identification algorithm is not available for mobile phone

Proposed Solution

Main idea:

A Method and Apparatus for photos' sensitive information identification on mobile phone is proposed,which include two parts: Image-based sensitive information identifierbased on the photo's context and features on mobile phone and automatic confidence-based tagging and altering decider.For sensitive image detection, we fetch a sensitivity score based on percentage of characters area from image itself and image context information such as neighbor image's sensitivity andfocus length to

judge image's sensitivity. Then we tag images with three sensitive levels based on above sensitive score and provide differentprotection policy for different level sensitive images.

A Strong observation lead an innovation:


•Photos taken by mobile phone may contain sensitive personal information, these low percentage (may less 10%) should be identified efficiently

•sensitive information Cannot be scatter Words, and words percentage on pic should be a very good indicator;

•Given sensitive pic distribution ,even false negative prediction didn't impact user experience(e.g. no more than 20 pictures in total if there are 500 pictures)
•Image-based sensitive information identification on mobile phone should not be too complex for power saving

•The context of the photos will help the information identification: photos continuously on mobile phone , near focus and etc.

1



Page 02 of 6

Implementation Details:

Part 1: Image-based sensitive information detection

In this part, we focus on obtaining a sensitive score based on percentage of characters area and image context information. As fig 2 shown, the below are specific steps. Step 1: Pre-processing and Binarization, The color images are first converted to grade-scaleimages, then binarize those grade-scale images with Canny edge method. After that, we also use the morphology operations to remove some noise edges. We can get a binary image after above processing.

Step 2: Indentify Suspicious Characte...