Anthony D. Wood

Research Projects

Projects to which I have made significant contributions are described below in a little more detail.

Enhanced Assisted-Living

Aging populations are stressing the U.S. healthcare system, and many are looking for ways to reduce costs while maintaining quality of care. Early detection of pathologies reduces the burden of disease treatment and improves outcomes, but it requires a long-term shift from a centralized, expert-driven, crisis-care model to one that permeates personal living spaces and involves informal caregivers, such as family, friends, and community.

AlarmNet is a wireless sensor network architecture and testbed for the independent- and assisted-living domains. It uses physiological, environmental, and activity sensors to monitor the health status of residents. Results of back-end analysis of behavior rhythms are fed back into the network for dynamic context-aware privacy and power management. Security of data and real-time query management is supported.

AlarmNet is the product of a large collaborative effort. My contribution to its design and development is in the following areas:

• System design and architecture. Includes front-end body networks, emplaced sensors, WSN–IP gateways, user interfaces, and back-end analysis programs. The architecture is heterogeneous, hierarchical, and open.
• System implementation in motes, stargates, and IP clients, including 80% of the nesC, C, and Java code.
• SenQ, a query system for efficient, real-time data extraction from in-network sensors. It is lightweight enough to co-reside with a graphical library on a mote-based embedded user interface.
• AMSecure lightweight link-level security for the WSN using the CC2420's built-in AES crypto support.

We've built a testbed that incorporates many sensor types (some shown above) and user interfaces, and continue to address difficult problems related to data association, power, privacy, security, pathology inference, and others.

More information is available on the AlarmNet site.

Secure Routing

Routing is an obvious target for attack in wireless networks that use multi-hop communication. One attack method is to corrupt the routing state to misdirect or disrupt traffic. Secure Implicit Geographic Forwarding (SIGF) is a family of configurable, secure routing protocols for WSNs. It builds on IGF, a completely state-free, dynamic geographic forwarding protocol.

SIGF comprises three protocols which extend IGF and populate the gap between pure statelessness and traditional shared-state security.

• SIGF-0 keeps no state, but uses nondeterminism and candidate sampling to achieve high packet delivery ratios probabilistically.
• SIGF-1 keeps local state, building reputations for its neighbors to aid in next-hop selection.
• SIGF-2 uses state shared with neighbors to provide the strongest defense against attack, yet at the greatest cost.

Each protocol encompasses the features of the previous, layering additional mechanisms to defend against more sophisticated attacks. The system can dynamically select the appropriate protocol and modify parameters for the best security vs. performance tradeoff.

SIGF won a best paper award (SASN '06), and is collaborative work with Tian He and Lei Fang.

Anti-Jamming

Denial-of-service from jamming is difficult to prevent with the limited resources available to most ad hoc and WSN devices. Radio transmission is an energy-expensive operation, yet an attacker can easily interfere with it. Using only software modifications to a compromised node, an attacker can efficiently jam the network.

I developed two solutions for dealing with the jamming problem at different layers:

• DEEJAM: Defeating Energy-Efficient Jamming in IEEE 802.15.4-based WSNs.

  DEEJAM operates at the MAC-layer, and uses four defensive mechanisms together to defeat or diminish the effectiveness of jamming by compromised sensors. Each additional defense addresses a different jamming attack. The end result is a novel protocol that allows network nodes to continue to operate—and communicate—in the presence of a jammer. DEEJAM was evaluated on an embedded platform, the MICAz mote, rather than in the idealized environment of a simulator, and appeared at SECON in 2006.

• JAM: Jammed-Area Mapping in WSNs

  

  Preventing denial-of-service attacks in WSNs is not always possible due to the limited resources available to network nodes and the ease with which attacks are perpetrated. Rather than jeopardize design requirements which call for simple, inexpensive, mass-producible devices, we propose a coping strategy that detects and maps jammed regions.

  JAM is a mapping protocol for nodes that surround a jammer which allows network applications to reason about the region as an entity, rather than as a collection of broken links and congested nodes. This solution is enabled by a set of design principles: loose group semantics, eager eavesdropping, supremacy of local information, robustness to packet loss and failure, and early use of results. It was published in RTSS 2003.

WSNs for Environmental Science

Wireless sensor networks for ecological research are often deployed in harsh environments where they must survive the elements of nature and function for extended periods of time with no further access for several months. Imagine a monitoring system deployed in polar regions where access is available only during certain months of the year. The system as a whole must function even when parts of it fail. The need for such a reliable, fault-tolerant system is the main motivation for the design of LUSTER—Light Under Shrub-Thicket for Environmental Research.

The system is designed to support:

• Fast deployment time.  Deploying large numbers of sensor nodes is time consuming. A self-organizing system is needed to free users from laborious setup tasks.
• Deployment assurance.  Remote deployments require that we have assurance that the system will continue to run without expensive and time-consuming return visits.
• Reliable and trusted data service.  The system must provide good sensor readings despite potential sensor malfunctions, node failures, and lossy communication links.
• Online data access.  Access is provided to historical and real-time data.
• Heterogeneity.  The system architecture and its query system may include many types of sensors. The query system supports any direct sample and split-phase sensors. An extensible API allows driver development for custom sensors where special handling of hardware or processing are required.
• Environmental challenges.  Rain and humidity are mostly tolerated through sealed packaging and dessicant inserts. However, some challenges are most unexpected. For example, we learned that one of the hazards for equipment left on the islands of the Eastern Shore of Virginia is that exposed cables are subject to the teeth of small critters.

I was a lead designer of LUSTER's multi-tier architecture, consisting of sensors, data stores, gateways, and back-end server. Sensors use the query system I originally developed for AlarmNet. I also collaborated in the design of LiteTDMA, a contention-free, energy-efficient MAC protocol tailored to the system's requirements.

In cooperation with environmental scientists, we deployed LUSTER on a barrier island off the Eastern Shore of Virginia. I led teams of students on three trips to the island to deploy and validate sensors, install infrastructure, and collect data. Each time, additional complexities were discovered that continue to drive our research in reliability.

RSS Feed Aggregation

Rnews is a server-side feed aggregator that we intend to remain lightweight and swift, with as few dependencies as possible. Why would you want one server-side? Because it is:

• Easy.  You interact with the aggregator with your web browser, where you're going to be reading your news anyway.
• Accessible.  Read news from anywhere you have a net connection, whether work, the beach house, or a coffee shop.
• Consistent.  What you've seen and what you haven't is kept consistent, no matter where you are.

Here's an excerpt of its features: it supports RSS and Atom feeds, has a step-by-step web install interface, supports multiple users, auto-imports feed info from the RSS feed, and has several view modes (all, category, list, single-feed). The full list is at the project homepage.

These features come with good performance. Feeds can be cached, reducing the burden on servers. A lightweight AJAX implementation improves usability and responsiveness. Of course, all pages are XHTML and CSS compliant.

Logins over SSL and cookies protected with HMAC-SHA1 protect your sessions. Passwords hashes are securely stored in the database to protect against compromise. A modified kses library strips unwanted tags from RSS feeds to mitigate XSS attacks.