Privacy against unsolicited radio-frequency sensing using machine learning

Liu, Zikun

Permalink

https://hdl.handle.net/2142/122184 Copy

Description

Title

Privacy against unsolicited radio-frequency sensing using machine learning

Author(s)

Liu, Zikun

Issue Date

2023-12-08

Director of Research (if dissertation) or Advisor (if thesis)

Vasisht, Deepak

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Wireless
• Embedded Systems
• Privacy, Sensing
• IoT
• RF
• Machine Learning

Abstract

In the last decade, both academia and industry have relied on passive radio frequency (RF) sensing to enable new capabilities for smart devices. Passive RF sensors can capture radio signal reflections from human bodies to track occupancy of rooms [1], motion patterns of occupants [2], and in more advanced systems, the breathing [3, 4], heart rate [5], sleeping patterns [6], keystrokes [7], and even emotions of occupants [8]. Recently, Google has incorporated such passive sensing into their smart home devices [9, 10] and Amazon received an FCC waiver [11] to conduct testing for the same. Notably, such sensors work without requiring users to carry a device and operate successfully through walls and other obstacles. Such tracking opens up a completely new set of privacy challenges. Smart devices can sense and mine behavioral data, passers-by can see if a home is empty, and neighbors can eavesdrop on your activities. Walls, curtains, and doors are meant to offer privacy to our indoor spaces. However, this notion of privacy no longer holds in the presence of RF-based sensing mechanisms that sense this information through walls. What makes this even more challenging is that it is near impossible for humans to evade such sensing. Human bodies naturally interact with radio signals and create small modifications that are then used for tracking. We propose a new framework for enabling privacy and user-control in the context of RF sensing – privacy by hallucination. We build new hardware-software techniques to inject fake ‘ghost’ reflections that appear like humans. By injecting false data and controlling false data injections, we can corrupt sensed information and allow users to regain control of private spaces. Part of the text is derived from [12].

Graduation Semester

2023-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/122184

Copyright and License Information

Copyright 2023 Zikun Liu

Owning Collections