Proceedings of the 5th workshop on Data management for sensor networks

Sensor networks aim to offer unprecedented means of monitoring the physical world, thus enabling entirely new applications. Many areas of science contribute to the research on sensor networks, for which reason many conferences exist that either cover or are devoted solely to sensor networks. The International Workshop on Data Management for Sensor Networks series, which was inaugurated in 2004 and of which this workshop is the fifth edition, stands out as a unique forum devoted to early and innovative work on data management in sensor networks. The workshop thus fills the gap in-between the database and other sensor network research areas. The scope of DMSN'08 covers all important aspects of sensor data management, including data acquisition, processing, and storage in remote wireless networks; the handling of uncertain sensor data; and the management of heterogeneous and sometimes sensitive sensor data in databases. The resource-constrained, lossy, noisy, distributed, and remote nature of wireless sensor networks implies that traditional database techniques often cannot be applied without significant retooling. Challenges associated with acquiring, processing, and archiving large-scale, heterogeneous sets of live sensor data also call for novel data management techniques. The inherently incomplete and noisy nature of sensor data further calls for techniques for data cleaning, inference, approximation. Finally, in many applications, the collecting of sensor data raises important privacy and security concerns that require new protection and anonymization techniques. DMSN'08 received 20 submissions, each of which was assigned to three or four members of the program committee. Based on the reviews and discussions among the program committee members, 8 papers were accepted for inclusion in these proceedings and for presentation at the workshop. The papers are grouped into three sessions, the first of which concerns in-network aggregation. With the purpose of reducing communication costs, Baljeet et al. study the use of tree topologies based on dominating sets for the forwarding and aggregation when computing MAX queries. Next, Cho et al. propose a new, so-called partial ordered tree that is capable of exploiting spatial correlation among sensor readings when performing top-k monitoring. Finally, Kontaki et al. propose a distributed solution for computing the d-hop k-data coverage query that generalizes previously considered queries. The second session concerns query processing trade-offs that involve the use of energy. In particular, Tang and Cao propose a data-driven power management framework, with which data accuracy and communication latency can be traded for improved energy efficiency. Trajcevski et al. study the trading of latency for improved balancing of energy consumption in a sensor network. The third session covers various aspects of complex sensor data processing. First, Mihaylov et al. consider the use of in-network joins for data integration in ad hoc networks and sensor and streams systems. Next, Evers et al. propose to associate sensor readings with time intervals rather than time points and then consider the use of two Hidden Markov Models, where a sensor value may extend across multiple hidden states, in this context. Their focus is on inference algorithms for these models. Karpinski and Cahill end this session by proposing a stream-based language targeted specifically at the programming of wireless networks encompassing both sensors and actuators. The workshop concludes with a panel. In this panel, well-known researchers in the community of sensor data management present and discuss specific new applications of sensor network technologies, as well as the notable technological challenges posed by these applications. As such, the panel contributes to setting new directions for the field.