I spoke on a panel today at the mid-year meeting of the American Bar Association’s Technology in the Practice and Workplace Committee.  The focus was on legal issues associated with RFID in the Workplace – other participants included representatives from the ACLU and various law firms representing either unions or companies that use RFID.  While the meeting was held at Seattle University, attendees included Professors from around Washington as well as lawyers and ABA members from across the country.

While I provided expertise on the existing attacks (and defenses) against existing RFID systems, I also emphasized privacy concerns with emerging RFID systems.  In particular, I explained how the RFID-based RTLS systems that are appearing in hospitals and government buildings can be abused.  Workers can be tracked either explicitly or through the equipment they use, patterns of association (e.g. relationships) between people can be made – as can changes in those patterns, and lots of unanticipated higher-level context can be extracted from raw sensor data long after it was recorded.  The union-side lawyer (Robert Lavitt from Schwerin Campbell Barnard Iglitzin & Lavitt LLP) also brought some interesting notes regarding past cases involving RFID tracking in hospitals. A key point in the privacy debate was on whether or not the benefit of such systems outweighed the privacy risk.  This is a recurring trade-off for context-aware computing technologies and it has been studied from many angles: “privacy vs. utility”, “privacy and proportionality”, “privacy vs. benefit”.  It’s more than likely that looming debates over emerging technology will form around this trade-off…

Overall the panel discussion was great and there should eventually be some references online, for now the day’s agenda provides an overview.

I presented the results of our first RFID Ecosystem measurement study today at MobiSys 2009 in Krakow, Poland. The study itself took place during October and November 2008 and included analysis of data from over 300 passive EPC Gen 2 RFID tags carried by more than 70 people.  A few of our high-level findings include:

(1) The total amount of data generated in 4 weeks was not that large – using a non-lossy compression algorithm, we shrunk 1.5 million tag read events to only 1.3 MB! However, there is more than an order of magnitude data blow-up after probabilistic filtering – a challenge for data processing but not storage.

(2) On average, tag detection rates were most influenced by type of tag and type of object.  This comes as no surprise to those familiar with passive RFID.  However, we did show that through “multi-tagging” (i.e., attaching multiple tags to an object at different orientations) we could dramatically improve detection rates for an object – by at least a factor of 2!  In addition, while overall detection rates were lower across the 4-week longitudinal study, there were huge differences across participants.  It turns out that the participants with the highest detection rates were also those who used our applications the most – suggesting that users could indeed understand and optimize RFID sensors if they were compelled to do so by useful applications.

(3) I also presented some details on our findings regarding privacy.  First, and perhaps surprisingly, there were no reported or detected privacy breaches.  Secondly, while users demanded sophisticated privacy controls before enrolling in our study, most created a only a small set of privacy rules on day one and never changed them – and only one user ever deleted any data!

Krakow is beautiful and there are quite a few other very interesting papers! – See the MobiSys 2009 technical program for the complete list. For the complete details on our work, please see our paper.

The Markovian Streams (MStreams) Project now has website: http://mstreams.cs.washington.edu/

Markovian streams are correlated, uncertain, ordered data streams; technically, they’re the output of probabilistic inference on a temporal graphical model.  The goal of the MStreams project is to develop algorithms that allow for efficient management and querying of Markovian streams using techniques from probabilistic data management.

This work is especially exciting from an RFID systems engineering perspective because we can generate and work with MStreams that represent RFID data – and we are.  In our Cascadia system we process RFID streams with particle filters to produce MStreams.  This process  transforms a raw RFID data stream (complete with missing tag reads and gaps from “dead zones”) into a stream containing a smoothed, probabilistic representation of location (i.e. a probability distribution over the person or object’s location at each timestep).  We’re now looking at adapting the MStreams Project’s database engine, Lahar, for use with our probabilistic RFID streams.  This should facilitate expressive queries over RFID data while improving accuracy through probabilistic data management.  More work in this vein soon…

I was fortunate to present our work on the RFID Ecosystem at the EMC Innovation conference in Franklin, MA today.  The conference marks the second annual gathering of EMC Corporation’s Innovation Network, a worldwide collaboration of advanced technology researchers exploring a variety of areas such as service-oriented infrastructures, web 2.0 storage, information-centric security, virtualization and information grids.

Our friends at RSA Labs (the Security Division of EMC) invited a talk on the RFID Ecosystem as one of three talks in a session on academic research.  In my talk I presented an overview of the RFID Ecosystem project and then drilled-down to describe some key challenges for pervasive RFID data management (e.g., uncertainty, privacy) and how we address them with the Cascadia system (my talk is available here).  The talk went quite well and was followed by some great questions and discussion with a few of the more than 1,000 attendees.  After the talk I had a chance to meet with researchers from RSA as well as ERC, EMC’s new research center in Beijing, China.  I received a lot of great feedback from these very interesting groups – the conference was a fantastic experience overall!

The Federal Trade Commission (FTC) and the University of Washington’s Law school are co-hosting a Town Hall meeting today at UW’s Law building.  The event is titled “Pay on the Go: Consumers and Contactless Payment” and features a series of panel discussions which will include representatives from government, banking, credit card companies, RFID industry consortiums, organizations for consumer rights and privacy as well as technical experts from industry and academia.  The goal of the meeting is to “examine the increasing prevalence of contactless payment devices in everyday consumer transactions, including credit card purchases and public transit use; consumer awareness and education initiatives regarding these developments; security and privacy threats and proposed solutions; and emerging technologies and practices that may shape the contactless payment marketplace over the coming years.”
A few notable participants include Prof. Kevin Fu from the University of Massachusetts, Amherst and our very own Prof. Tadayoshi Kohno and Prof. William Covington.  Should be a fascinating discussion, watch the streaming video of the Pay on the Go Town Hall Meeting here!

The Journal of the American Medical Association published a report on Wednesday that showed certain RFID systems may induce hazardous incidents in nearby critical care medical equipment.  The study examined the effects of a 128 kHz active RFID system and a 868 MHz passive RFID system on 41 different types of medical devices.  Out of 123 tests, 34 led to electromagnetic interference incidents, 22 of which were classified as hazardous, 2 as significant, and 10 as light.

Though many consider the result of this study to be a serious and shocking failure in engineering, some RFID vendors claim that the problem could be avoided by lowering the power output of the RFID readers.  In any case, this study highlights a key obstacle for pervasive RFID systems: it will probably not be possible to deploy RFID readers in all locations of interest due to a variety of issues including electromagnetic interference, budgetary issues, or aesthetic concerns.  As such, a successful RFID system will have to infer when a tag enters a location which is not covered by an RFID reader.  This is one of the challenges which we designed Cascadia, the RFID Ecosystem’s event detection middleware, to address.  In particular, Cascadia allows an admin to define a graph that describes locations of interest (which may or may not be covered by an RFID reader – see image below).  Cascadia then tracks tags over this graph and can infer with some probability when a tag enters a location that is not covered by an RFID reader.  See our MobiSys 2008 paper for details!

The Guardian reports that London’s Oyster transit card may have been hacked by a group of Security researchers at Radboud University in Holland.  The hack allows an attacker to clone the Oyster card’s RFID chip after cracking its encryption in just a few seconds.  This is the latest in a series of Mifare smartcard hacks which have compromised European transit card systems in the last year.

It’s interesting to note that the authorities are not considering the hack to be a serious threat.  London Transport has claimed not only that they can detect fraudulent use within 24 hours using checks in software (a level of security often ignored by hardware hackers), but that a criminal could gain at most about £3 per cloned card.  As such, the incentive to clone Oyster cards probably isn’t that great.  One point which the article does not highlight and which is perhaps more concerning from a privacy standpoint, however, is that London Transport collects and can process data on 38 million journeys per week to identify individual instances of fraudulent use (and who knows what else?).

We just returned from two weeks at the SIGMOD and MobiSys conferences where we presented demos of the RFID Ecosystem’s event detection infrastructure, Cascadia. Cascadia allows developers and end users to declaratively specify meaningful high-level events (e.g. “a nurse has entered patient X’s room with equipment Y”, “I leave the building without my car keys”) using a SQL-like sequence language or with an intuitive GUI called Scenic. Cascadia can then continuously extract these events over a stream of incoming, uncertain RFID data using probabilistic data management techniques. The Cascadia demos at SIGMOD and Mobisys illustrated Cascadia’s operation from end-user event specification, to event detection, to event notifications for one of two demo applications. The original demo proposals are available on our publications page; both demos are also entirely web-based and will be posted online sometime in the coming weeks.

A group of delegates from various Korean RFID associations visited UW CSE today to learn about the RFID Ecosystem project, to discuss challenges, and to share findings on future scenarios for RFID use. We shared demonstrations, presentations, and experience over the course of the morning and found that we share many of the same challenges (e.g. uncertainty in sensor data, system scalability, user privacy). The 20 RFID experts represent Korea’s large and growing interest, investment, and innovation in RFID technology. Korea’s focus on RFID is driven forward by projects like New Songdo City, a planned “U-city” and international business center which will include many elements of ubiquitous computing infrastructure including, possibly, RFID readers deployed throughout the city. In this scenario, RFID would be used to promote recycling of tagged products as well as possibly for health care applications; whether the ubiquitous RFID deployment becomes reality or not, the $25 billion project will include a $297 million RFID research center when completed. After their visit the delegates continued on to the RFID World conference in Las Vegas.