Tristan Henderson

CRAWDAD: A Community Resource for Archiving Wireless Data at Dartmouth

Wireless network researchers are seriously starved for data about how real users, applications, and devices use real networks under real network conditions. CRAWDAD (Community Resource for Archiving Wireless Data at Dartmouth) is a new National Science Foundation-funded project to build a wireless-network data archive for the research community. It will host wireless data and provide tools and documents to make collecting and using the data easy. This resource should help researchers identify and evaluate real and interesting problems in mobile and pervasive computing. To learn more about CRAWDAD and discuss its direction, about 30 interested people gathered at a workshop held in conjunction with MobiCom 2005.

Privacy in Location-Aware Computing Environments

This study explores how privacy preferences vary with place and social context. These findings are useful for designing privacy policies and user interfaces for pervasive computing.

Less is more: energy-efficient mobile sensing with senseless

We present SenseLess, a system that leverages the different energy consumption characteristics of sensors to maximise battery life in mobile-sensing applications. We use the less expensive sensors more often, thereby enabling us to use the more expensive sensors less frequently. In the context of location-aware services, experimental results indicate that for a typical indoor and outdoor walk, compared to a simple GPS-based system, our SenseLess system can reduce energy consumption by more than 58% when determining a users location, while maintaining the fidelity of the sensed data. This extends the battery life of a typical handheld device from 9 hours to 22 hours.

Modelling user behaviour in networked games

In this paper we attempt to gain an understanding of the behaviour of users in a multipoint, interactive communication scenario. In particular, we wish to understand the dynamics of user participation at a session level. We present wide-area session level traces of the popular multiplayer networked games Quake and Half-Life. These traces were gathered by regularly polling 2256 game servers located all over the Internet, and querying the number of players present at each server and how long they had been playing. We analyse three specific features of the data: the number of players in a game, the interarrival times between players and the length of a players session. We find significant time-of-day and network externality effects in the number of players. Player duration times fit an exponential distribution, while interarrival times fit a heavy-tailed distribution. The implications of our findings are discussed in the context of provisioning and charging for network quality of service for multipoint and multicast transmission. This work is ongoing.

Exploiting Self-Reported Social Networks for Routing in Ubiquitous Computing Environments

Mobile, delay-tolerant, ad hoc and pocket-switched networks may form an important part of future ubiquitous computing environments. Understanding how to efficiently and effectively route information through such networks is an important research challenge, and much recent work has looked at detecting communities and cliques to determine forwarding paths. Such detected communities, however, may miss important aspects. For instance, a user may have strong social ties to another user that they seldom encounter; a detected social network may omit this tie and so produce sub-optimal forwarding paths. Moreover, the delay in detecting communities may slow the bootstrapping of a new delay-tolerant network. This paper explores the use of self-reported social networks for routing in mobile networks in comparison with detected social networks discovered through encounters. Using encounter records from a group of participants carrying sensor motes, we generate detected social networks from these records. We use these networks for routing, and compare these to the social networks which the users have self-reported on a popular social networking website. Using techniques from social network analysis, we find that the two social networks are different. These differences, however, do not lead to a significant impact on delivery ratio, while the self-reported social network leads to a significantly lower cost.

CRAWDAD: a community resource for archiving wireless data at Dartmouth

Wireless network researchers are seriously starved for data about how real users, applications, and devices use real networks under real network conditions. CRAWDAD, a Community Resource for Archiving Wireless Data at Dartmouth, is a new NSF-funded project to build a wireless network data archive for the research community. We host wireless data, and provide tools and documents to make it easy to collect and use wireless network data. We hope that this resource will help researchers identify and evaluate real and interesting problems in mobile and pervasive computing. This report outlines the CRAWDAD project, the kick-off work-shop that was held at MobiCom 2005, and the latest news. how real users, applications, and devices use real networks under real conditions, and how mobile users actually move about. This data helps us to identify and understand the real problems, to evaluate possible solutions, and to evaluate new applications and services.

IRONMAN: Using Social Networks to Add Incentives and Reputation to Opportunistic Networks

Opportunistic networks enable users to communicate in the absence of network infrastructure. But forwarding messages in such a network incurs costs for nodes in terms of energy and storage. This may lead to nodes being selfish and not forwarding messages for other nodes, resulting in degraded network performance. This paper presents a novel incentive mechanism for opportunistic networks that uses pre-existing social-network information to detect and punish selfish nodes, incentivising them to participate in the network. Trace-driven simulations demonstrate that our mechanism performs better than existing mechanisms, and that social-network information can also be used to improve existing incentive mechanisms.

What online gamers really think of the Internet

The Internet has provided a network infrastructure with global connectivity for the games industry to develop and deploy online games. However, unlike the document interface paradigm of the World Wide Web (WWW), these online games have more stringent requirements that are not fulfilled by the Internets best effort service model.A key characteristic of online games is the possibility of having multiple participants share the same experience. Consequently, the volatile nature of the Internet can affect the enjoyment of all, or at the very least a few, of the users. To ameliorate the impact caused by network problems that may arise during game play, game developers have adopted adaptation techniques in the design and implementation of online games. However, little is known of how the user perceives these mechanisms.This paper presents the results of a questionnaire targeted at the online gaming community to provide insight into what users really think of the Internet and its impact on their playing experience. One of the main results is to demonstrate that the existing mechanisms fail to maintain the utility of the game at all times, leading to frustration on the part of the users. In spite of this, users are not willing to pay for any service guarantees.

Privacy-enhanced social-network routing

An opportunistic network of mobile nodes can be created when mobile devices work together to create an ad hoc store-and-forward architecture, with messages forwarded via intermediary encountered nodes. Social-network routing has been proposed to route messages in such networks: messages are sent via nodes in the senders or recipients friends list. Simple social-network routing, however, may broadcast these friends lists, which introduces privacy concerns. This paper studies mechanisms for enhancing privacy while using social-network routing. We first present a threat analysis of the privacy risks in social-network routing, and then introduce two complementary methods for enhancing privacy in social-network routing by obfuscating the friends lists used to inform routing decisions. We evaluate these methods using three real-world datasets, and find that it is possible to obfuscate the friends lists without leading to a significant decrease in routing performance, as measured by delivery cost, delay and ratio. We quantify the increase in security provided by this obfuscation, with reference to the classes of attack which are mitigated.

Map: a scalable monitoring system for dependable 802.11 wireless networks

Many enterprises deploy 802.11 wireless networks for mission-critical operations; these networks must be protected for dependable access. This article introduces the MAP project, which includes a scalable 802.11 measurement system that can provide continuous monitoring of wireless traffic to quickly identify threats and attacks. We discuss the MAP system architecture, design decisions, and evaluation results from a real testbed.