PlaceAvoider: Steering First-Person Cameras away from Sensitive Spaces

Abstract

Abstract-Cameras are now commonplace in our social and computing landscapes and embedded into consumer devices like smartphones and tablets. A new generation of wearable devices (such as Google Glass) will soon make 'first-person' cameras nearly ubiquitous, capturing vast amounts of imagery without deliberate human action. 'Lifelogging' devices and applications will record and share images from people's daily lives with their social networks. These devices that automatically capture images in the background raise serious privacy concerns, since they are likely to capture deeply private information. Users of these devices need ways to identify and prevent the sharing of sensitive images.As a first step, we introduce PlaceAvoider, a technique for owners of first-person cameras to 'blacklist' sensitive spaces (like bathrooms and bedrooms). PlaceAvoider recognizes images captured in these spaces and flags them for review before the images are made available to applications. PlaceAvoider performs novel image analysis using both fine-grained image features (like specific objects) and coarse-grained, scene-level features (like colors and textures) to classify where a photo was taken. PlaceAvoider combines these features in a probabilistic framework that jointly labels streams of images in order to improve accuracy. We test the technique on five realistic firstperson image datasets and show it is robust to blurriness, motion, and occlusion.

Privacy behaviors of lifeloggers using wearable cameras

“…Previous work proposed algorithms to detect photos taken in specific 'sensitive spaces' [27], but more research is needed to determine how automated mechanisms can incorporate location attributes. Indoor versus outdoor photo classification has been studied in the computer vision literature and may provide a starting point for automated image analysis [4] as our results indicate people share fewer indoor images (but a large fraction are shared nevertheless).…”

“…Recent work has shown how opportunistic images generate new threats to users, such as allowing 3-D models of their environment to be surreptitiously created and enabling 'virtual theft' [28]. Complementary work includes defensive frameworks in which users define policies based on physical location, so that photos taken in predefined sensitive spaces can be recognized and then deleted or quarantined for review [27]. Researchers have also qualitatively studied reactions of bystanders to wearable camera devices [10,20], as well as the sensitivity of lifelogging data and how it could be automatically altered to enable privacy-preserving processing [30].…”

To Log or Not to Log? SWOT Analysis of Self-Tracking

“…Recently, researchers developed a technique called PlaceAvoider for owners of first person cameras to 'blacklist' sensitive spaces (like bathrooms and bedrooms). This technique provides a way to identify and prevent the sharing of sensitive images (Templeman, Korayem, Crandall, & Kapadia, 2014).…”

Reactive security

“…Practical ABAC systems will thus require identifying attributes that can be extracted quickly and accurately, while still providing sufficient discrimination for access control decisions. Our PlaceAvoider system [34] is one initial attempt at this trade-off: that approach is ABAC-based, but relies on a single attribute (image location) that could be recognized accurately through a combination of GPS and automatic image recognition techniques.…”