Privacy-preserving Image Processing with Binocular Thermal Cameras

Abstract

Today, cameras and digital image processing are transforming industries and the human environment with rich, informative sensing. However, image processing is not utilized nearly as much in homes where concerns about image privacy dominate. In a preliminary study with 200 participants, we found 21% would reject a camera based system even if the system was designed to not report images as they could still be collected if the camera system was hacked. In this paper, we demonstrate a hardware-based approach for privacy-preserving image processing: the ability to automatically extract information from imaging sensors without the risk of compromising image privacy, even if the system is hacked. The basic idea is to limit both the memory available on board the camera and the data rate of camera communication to prevent a full image from ever being extracted. As a proof of concept, we prototype a system, called Lethe, that tracks and identifies individuals by height with a thermal camera as they move from room to room. Our results show that Lethe can detect the presence of individuals with 96.9% accuracy and determine their direction of travel with 99.7% accuracy. Additionally, Lethe can identify individuals 96.0% of the time with a 5cm (~2in) or greater difference in walking height and 92.9% with a 2.5cm (~1in) or greater difference. Finally, Lethe performs this processing with only 33 bytes of memory (or 0.69% of the full thermal image).

Experimental Evaluation of Reduction Technique for Time Delay of Turning Off Light to Occupancy Lighting Control

“…Because a privacy‐preserving image processing has been developed as shown in Ref. [20], the discussion of the privacy preserving is out of scope of this paper.…”

Using Convolutional Neural Networks with Multiple Thermal Sensors for Unobtrusive Pose Recognition

“…There has been a concern highlighted that even if a system was not designed to report images, the images could still be accessed if the system was hacked. The study in [21] proposed an approach to detecting and identifying people while the captured data preserved their privacy. The system, called Lethe, used limited memory storage and data transmission so that full thermal images would not be compromised in the event of the system being hacked.…”

“…In contrast to the previously highlighted privacy preserving approaches, our work utilises the low-resolution images captured by TISs. High-resolution thermal images have proven to be advantageous while preserving privacy [20,21] where the facial features of a person were successfully hidden while maintaining a high-quality image for the rest of the scene. A high resolution was particularly practical in [21] as identifying the pixel representing the top of a person's head allowed for the eventual estimation of their height, therefore, facilitating the identification of the person.…”

“…High-resolution thermal images have proven to be advantageous while preserving privacy [20,21] where the facial features of a person were successfully hidden while maintaining a high-quality image for the rest of the scene. A high resolution was particularly practical in [21] as identifying the pixel representing the top of a person's head allowed for the eventual estimation of their height, therefore, facilitating the identification of the person. A low-resolution image would not allow for an accurate determination of the top of the head, subsequently hindering the system's capability to use height to accurately differentiate between people.…”

Profiling and improving the duty-cycling performance of Linux-based IoT devices

“…Such applications benefit from edge and fog computing to improve responsiveness and minimize the overhead of data exchange with cloud platforms [15]. For example, image processing through artificial intelligence on IoT devices presents several benefits including faster decision-making and lower reaction time to environmental changes, less wireless interference with nearby devices, and improved security due to eliminating the need to upload raw images to a cloud service [16]. According to ABI Research [17], 90% of the data generated by edge devices is being processed locally.…”