Recent Events on Security, Privacy and Machine-Learning

The TerraSwarm Security and Privacy Meeting, September 16-17, was hosted by United Technologies at the Rochester Institute of Technology, Rochester, NY. The event attracted more than 30 attendees, including faculty and students from five universities (Caltech, UC Berkeley, UCSD, University of Michigan, University of Pennsylvania) as well as industry sponsors.

Participants were exposed to a mix of presentations by speakers from industry and academia, interactive demonstrations from RIT security personnel, and open discussions on topics of security and privacy for TerraSwarm applications. The breakout sessions identified threat models, security and privacy goals, and advances required to achieve these goals at all levels of the swarm architecture.

Preceding the Security & Privacy Meeting, a one-day Machine Learning Hands-On Workshop was offered. The goal of this hands-on workshop was to familiarize attendees with the work being done in Machine Learning within TerraSwarm. New faculty member, Jeff Bilmes (UWash) provided an overview of the Machine Learning Tools available. Ilge Akkaya (UCB) provided an overview of Ptolemy models for Machine Learning and Optimization for the swarm. After presentations, teams of 2 to 4 developed Machine Learning applications using various components of the SwarmOS.

Prabal Dutta Named one of the 'Brilliant Ten' by Popular Science Magazine

TerraSwarm faculty member Prabal Dutta was recently named one of the '2014 Brilliant Ten' by Popular Science Magazine.  Their annual list predicts  "tomorrow's Einsteins, Zuckerbergs and Marie Curies".

Popular Science Magazine explains why he was selected:

Dutta creates tiny sensors that scavenge energy from their surroundings so that they can run forever, ushering in the Internet of Things.

A University of Michigan article provides further detail about his work:

Dutta's technologies scale down to the very small. His architectures are enabling millimeter scale computers, among the tiniest in the world. So-called smart dust could one day be embedded all around us, helping us monitor our environment.

Dutta designs hardware and software for so-called smart dust and larger sensors that don't need batteries because they can harvest energy from the world around them. They can run on power from the light in a room, the magnetic field around an electrical wire, or the heat from a shower head, for example.

To read more about Prabal Dutta's work, go to:

• http://www.popsci.com/article/science/brilliant-ten-2014
• http://ns.umich.edu/new/releases/22385-popular-science-puts-u-m-engineer-on-brilliant-ten-list

UPenn student Zhihao Jiang earns "Best in Session" Award at TECHCON 2014

University of Pennsylvania doctoral student Zhihao Jiang, was recently awarded Best in Session at SRC’s TECHCON 2014 in the Session on System Design Tools for his paper entitled Integrated Functional and Formal Modeling for Closed-loop Evaluation of Medical Device Software. Co-authors of the paper were Miroslav Pajic and Rahul Mangharam.

Their paper introduces a reconfigurable closed-loop testing platform for implantable cardiac devices which is part of their model-based software design framework.

According to co-author Rahul Mangharam: "Software accounts for the most rapid increase in reasons for recalls today.  Tools for rapid safety certification for software in life-critical medical devices are critical to the semiconductor industry. Speedy certification lowers the cost and barriers of entry in the high-margin medical devices industry. "

To read Jiang's award-winning paper or view his associated slide presentation or poster go to http://www.terraswarm.org/pubs/358.html

UPenn's new xLab uses "Internet of Things" to Design Interactive Entertainment

TerraSwarm PI Rahul Mangharam (Penn Engineering) has joined forces with faculty in Penn's School of Design to launch a new Experience Design and Technology Lab—  xLAB for short.  The goal of the collaboration is to use the "Internet of Things" to devise and deliver richer, more interactive entertainment experiences.  A Penn Current online article highlighted the new lab:

"xLAB began through collaborations between students in their respective classes through Penn’s Integrated Product Design and Embedded Systems programs. Students in xLAB brainstorm projects related to the future of entertainment systems, coupling digital content with internet-connected physical devices. They actively interact with design consultants, involve users to narrow down their needs, and then work together to build, refine, and test prototypes of those ideas."

According to Mangharam:

"The fundamental problem we are attacking is that the content delivery infrastructure today (e.g. cable TV) does not know anything about the context of the content. We are enabling new ways of content pacing and sequencing to personalize media consumption and develop the network architecture for context exchange between the content and the physical world with such content-coupled IoT." 

Examples of some of the technology being developed at xLab include an intelligent yoga mat, a vest that enables viewers to actually feel the action in a film, and interactive children’s toys.  xLab's introductory video highlights a few of these new technologies: http://www.upenn.edu/pennnews/current/2014-07-17/latest-news/xlab-designs-next-generation-immersive-experiences

Alex Halderman's Team at University of Michigan Study Traffic Light Security

New TerraSwarm faculty member, Alex Halderman and his team at the University of Michigan,  made news with their recent paper "Green Lights Forever: Analyzing the Security of Traffic Infrastructure," presented at the Workshop on Offensive Technologies (WOOT) at USENIX Security 2014 in August. Their research found that the state's traffic light infrastructure is relatively easy to hijack.  With the permission of the Michigan Traffic Authority, they were able to take control of a system of nearly 100 wireless networked traffic lights and successfully change the light commands.

According to their research:

"With the appropriate hardware and a little effort, [a hacker] can execute a denial of service attack to cripple the flow of traffic in a city, cause congestion at intersections by modifying light timings, or even take control of the lights and give herself clear passage through intersections."

Hacking the traffic lights is possible from half a mile away with nothing more than a laptop and wireless card that operates on the same 5.8-gigahertz frequency as the traffic lights.  Unencrypted wireless signals, default usernames and passwords and the use of a traffic controller were the primary reasons they were able to hack into the lights so easily.

Team members concluded:

"While other deployments may use different wireless radios or even wired connections between intersections we have no reason to believe there are any fundamental differences between the network we studied and other traffic signal systems. We believe that many traffic infrastructure devices created by various vendors and installed by various transportation departments will have similar security properties due to a lack of security consciousness in the entire field."

To read more, a simple google search will yield numerous articles about this research.
The three articles quoted for this blog post are listed below:

• The Register: Need a green traffic light all the way home? Easy with insecure street signals, say researchers
• NBC News: Traffic Lights Are Vulnerable to Simple Hack, Researchers Find
• TechTimes: Hacking into traffic lights with a laptop is apparently scarily easy

Call for Papers: Special Issue on the "Cyber‐Physical Systems for Medical Applications" of IEEE Design & Test

A Special Issue on the "Cyber‐Physical Systems for Medical Applications" of IEEE Design & Test is being planned for publication in September/October 2015. We invite you to submit your papers by October 15, 2014. Information about this Special Issue can be found at:
http://www.eng.ucy.ac.cy/theocharides/ieeedt/CFP_MedCPS2014_DandT.pdf

Prospective authors should follow the submission guidelines for IEEE Design & Test. All manuscripts must be submitted electronically to the IEEE Manuscript Central Web site at http://www.manuscriptcentral.com/. Indicate that you are submitting your article to the special issue on “Cyber‐Physical Systems for Medical Applications”. All papers will undergo the standard IEEE Design & Test review process.

For any question, feel free to contact the Guest Editors:
Paul Bogdan, pbogdan@usc.edu
Rahul Mangharam, rahulm@seas.upenn.edu

TerraSwarm Investigator Carlos Guestrin Talks about GraphLab's New Machine-Learning Software

TerraSwarm investigator and GraphLab co-founder / CEO Carlos Guestrin, was recently interviewed by gigaom.com about the release of their new machine learning commercial software, GraphLab Create.   GraphLab was launched in 2013 to develop an open-source software platform for high-performance, scalable machine learning and graph analytics.  The new software they have developed allows users to run a number of popular machine learning tasks on data stored in graphs or tables.

The article quotes Guestrin:

The goal of Create is to help savvy engineers or data scientists take their machine learning projects from idea to production. It includes a handful of modules for building certain types of popular workloads, including recommendation engines, graph analysis and clustering and regression algorithms.

One of Create’s main benefits is simplicity. Users write, test and deploy their jobs in Python, and the jobs execute in GraphLab’s C++ engine to step up the speed. Jobs can execute on a laptop or across a cluster of servers running Hadoop (with YARN), and built-in management tooling lets users monitor running jobs. 

The way iOS did for mobile applications, Guestrin wants GraphLab Create to make it so anybody with the inclination and a modicum of coding skills can start building machine learning applications without worrying about all the hairy details around optimization, deployment and other historically time-consuming processes. 

Can a software product like GraphLab Create democratize machine learning?  To read more of Guestrin's comments and what capabilities GraphLab plans to include in future versions of their software, go to: http://gigaom.com/2014/07/15/graphlab-wants-to/.