Privacy-aware image classification and search

Abstract

Modern content sharing environments such as Flickr or YouTube contain a large amount of private resources such as photos showing weddings, family holidays, and private parties. These resources can be of a highly sensitive nature, disclosing many details of the users' private sphere. In order to support users in making privacy decisions in the context of image sharing and to provide them with a better overview on privacy related visual content available on the Web, we propose techniques to automatically detect private images, and to enable privacy-oriented image search. To this end, we learn privacy classifiers trained on a large set of manually assessed Flickr photos, combining textual metadata of images with a variety of visual features. We employ the resulting classification models for specifically searching for private photos, and for diversifying query results to provide users with a better coverage of private and public content. Large-scale classification experiments reveal insights into the predictive performance of different visual and textual features, and a user evaluation of query result rankings demonstrates the viability of our approach.

Analyzing and Predicting Privacy Settings in the Social Web

“…This goal is similar to the work presented in [32], although addressing a different media type. In their work, Zerr et.…”

Automatic Privacy Classification of Personal Photos

“…Furthermore, the results from our feature analysis match their findings regarding the influence of persons and locations on u sers' privacy settings. Zerr et al [11] targeted privacy-aware image search with a Support Vector Machine (SVM) classifier trained on publicly available photos from Flickr. They used five visual features: faces, hue histogram, edge-direction coherence vector [7], SIFT features [5], and average brightness and sharpness.…”

“…However, in both projects [10,11], the "private" photos had been published on the web by Flickr users. Hence, these pictures were not considered private by their owners.…”

“…Moreover, users may want to share their photos online on social platforms like Flickr or Facebook. To support users in revealing their photos only to intended audiences, applications need to become more privacy-aware, as recommended in related research [1,10,11]. To achieve this, applications should be informed about the user's intended privacy setting, ideally for each photo individually [1].…”

Entity-Based Opinion Mining from Text and Multimedia

“…For our privacy classification experiments [49,50], we created a dataset of 90,000 "recently uploaded" images from Flickr with a minimum of 5 English tags. In order to create ground-truth, we created a social annotation game and used crowdsourcing to get the opinions of multiple individuals.…”