As our reliance on smart devices continues to grow, so does our exposure to potential privacy breaches. From smart fridges, delivery drones to robotic vacuum cleaners, these devices utilize vision to interact with their environment, inadvertently collecting and storing visual data about our lives.
In a groundbreaking bid to safeguard privacy, researchers from the Australian Centre for Robotics at the University of Sydney and the Centre for Robotics (QCR) at Queensland University of Technology have developed an innovative method of designing cameras. Instead of capturing and storing a complete visual representation, these cameras process and scramble the visual information before it is digitized, making the images indecipherable to the point of anonymity.
All devices connected to the internet, or the so-called "internet-of-things," uniquely include these newly termed sighted systems. However, these systems come with vulnerabilities. They can be hacked or misused to steal customer data, sometimes even with malicious aims. Now, these devices, which include the smart vacuum cleaners, are able to distort images while still being able to do their tasks – without compromising privacy.
Adam Taras, who contributed to the research as part of his Honours thesis, had remarkable insights regarding this issue. "While smart devices usher in changes to our lifestyles and workplaces, they should not become tools for surveillance. These devices don't require the same type of visual access as humans do. They do their functions based on a narrow spectrum of visual cues, such as color and pattern recognition," said Taras.
In their novel approach, the researchers demonstrated that the picture processing can be compartmentalized to a level beyond the purview of potential attackers. Dr. Don Dansereau, Taras' supervisor, said that unlike previous methods which obfuscated the images within the camera's computer rendering them attack-prone, their approach protects the integrity of images to a greater extent.
The team brought forward an open challenge to the wider research community, inviting others to attempt hacking their method, confident that the images couldn't be reconstructed in a recognizable format. "If these images were accessed by a third party, they wouldn't make sense, thereby preserving privacy," Taras added.
According to the researchers, this development is particularly important as more devices come equipped with built-in cameras. They see potential applications in sectors where privacy is paramount, including schools, hospitals, factories, warehouses, and airports. The team is hopeful that their work will inspire the development of more privacy-considerate AI tools.
Disclaimer: The above article was written with the assistance of AI. The original sources can be found on ScienceDaily.
