Division of Computer and Network Systems (CNS)

The publication of image data captured by ubiquitous surveillance devices, such as traffic cameras and security surveillance cameras, would greatly benefit various communities and enable many applications. However, sharing image data with untrusted parties would raise privacy concern due to potential sensitive content, like identities and activities that may be in the images. Standard image obfuscation techniques, such as pixelation and blurring, do not provide effective privacy preservation for people or objects represented in the data.

Many online platforms use economic mechanisms to estimate the best ways to match consumers and businesses with products and services. Effective matches may require using personal consumer data but doing so may intrude on consumers' privacy. This project will use formal concepts of privacy to analyze the use of personal information in mechanism design.

This project integrates advances in data science and key findings from psychological research to improve the prediction of trust in information on social media by modeling the psychological phenomenon known as the memory illusion. The memory illusion refers to memory errors that people make to remember information as an outcome of interpreting and making inferences from their past experience. This project will use social media data to examine the memory illusion with online information, and to understand how it is associated with people's trust in information on social media.

Due to the prohibitive costs of semiconductor manufacturing, most computer chip design companies outsource their production to offshore foundries. As many of these chips may be manufactured in environments of limited trust, problems of the piracy of intellectual property (IP) and the overproduction of integrated circuits (ICs) have emerged in recent years. This project focuses on designing a secure logic locking technique to enable protection against untrusted IC manufacturing. The developed solution is resistant to all known attacks.

Recent data breaches have exposed billions of user passwords to the dangerous threat of an offline password attacker who attempts to guess each user's password by brute force. Because an offline attacker can validate each password guess by itself using stolen password hashes from a data breach it is not possible to "lock out" an offline attacker after several incorrect guesses. The attacker is limited only by the computational resources necessary to mount a brute-force attack.