Efficient Monitoring Method of Personally Identifiable Information on Images Exposed in the Web

Abstract

Although retrieval systems of a personally identifiable information (PII) exposed in a text have developed rapidly, but retrieval method of PII exposed in an image is a challenging task. This paper proposes an efficient PII retrieval method that performs classification of image including PII. In the proposed method, the color, texture and shape-related features from images with PII are extracted by the gray-level co-occurrence matrix (GLCM) analysis. Then, our method adopts the multiple classifiers that are histogram, features, template matching, and support vector machine (SVM)-based image classifiers to classify the images with PII. The experimental results obtained using our method show a classification rate of 82% and an execution time of approximately 0.17 seconds per image for classifying images with PII. The method can be effectively applied to systems for personal information retrieval and exposure system on the web or privacy incident response system.