Past Research Projects
-
[DNC] Privacy Protection from Prying Eyes for Wearable Devices
Smartphones and tablets are excellent point-and-shoot cameras, with users taking dozens of pictures and videos anytime and everywhere. Wearables like smart goggles can record media in both public and private places, with little or no awareness from the subjects in the surroundings. The pervasive use of these devices can compromise the privacy of all the individuals that are unaware subjects of these pictures and videos, which could also be published without explicit consent on the Internet and social media. This project proposes a novel technology that removes unwilling subjects from media that includes them at capture time.
-
[P3] Privacy-Preserving Photo Sharing
Photo sharing services, e.g. Facebook and Flickr, are getting more popular nowadays, however, there are no protection of users' photos against the photo sharing service providers (PSPs). We are just forced to trust PSPs. In this work, we develop a photo encryption algorithm that can preserve privacy against PSPs while still maintaining the useful processing services provided by them, e.g. image scalability and quality enhancement. We also built a prototype transparently worked with Facebook. (NSDI 2013, app)
-
[Medusa] Programming Crowd-Sensing Task
Design and implementation of a high level programming language and associated runtime for crowd-sensing applications. The framework makes two strands of topics extremely easy to program, a) tasking smartphones and b) dealing with human mediation and incentives. (MobiSys 2012, code, demo video)
-
[MultiSensing] Multi-Processor Sensing
Attempt to answer the following question: In order to enable continuous personal sensing applications on mobile devices, how should we use ultra low power processor(LP) with an existing application processor(AP)? Investigate representative app components and explore the placement strategy between AP and LP that brings minimum energy consumption. (with MSR Redmond SERG group, Ubicomp'12)
-
[Odessa] Enabling Perception Applications on Mobile Devices
Partitioning mobile perception applications across mobile devices and the cloud infrastructure focusing on the performance. Design and implementation of lightweight, i.e. sub-second decision granularity, dynamic decision engine and associated runtime that can adapt to input variability, network and device heterogeneity in real-time. (with Intel Labs Seattle, MobiSys'11)
-
[SALSA] Energy-Efficient Network Interface Selection
Design and implementation of SALSA (Stable and Adaptive Link Selection Algorithm) for high data-rate and delay-tolerant mobile applications. Our algorithm intelligently defers the transmission opportunity to the future based on Lyapunov optimization framework anticipating that more energy-efficient WiFi AP, rather than 3G, would be avaiable soon. (MobiSys'10)
-
[UrbanTomography] Automatic Video Uploading System
We built an Urban Tomography system, which collects high resolution videos using smartphones and automatically uploads them to an Internet-connected server without any user interventiton. When transfer, the system deals with an automatic AP management issue, e.g. chunking data, network profiling for proper AP selection, blacklisting, etc. It had deployed to the LAX international airport for more than a year and the course projects in USC and UCLA. (JUT'10, JPER'09, [poster])
| Initially created: Oct. 8th, 2007 , Last Modified: Feb. 3rd, 2016 |
|