An UBIQUEST system provides a high level programming abstraction for rapid prototyping of heterogeneous and distributed applications in a dynamic environment. Such a system is perceived as a distributed database and the applications interact through declarative queries including declarative networking programs (e.g. routing) and/or specific data-oriented distributed algorithms (e.g. distributed join). Case-Based Reasoning is used for optimization of distributed queries when as there is no prior knowledge on data (sources) in networking applications, and certainly no related metadata such as data statistics.

Wireless sensor networks (WSNs) are composed of a large number of low-cost, low-power, and multi-functional sensor nodes that communicate at short distance through wireless links. They are usually deployed in an open and uncontrolled environment where attackers may be present. Due to the use of low-cost materials, hardware components are not tamper-resistant and an adversary could access a sensor's internal state. An adversary can easily capture even a single node and inserts duplicated nodes at any location in the network. If no specific detection mechanisms are established, the attacker could lead many insidious attacks such as subverting data aggregation protocols by injecting false data, revoking legitimate nodes and disconnecting the network if the replicated nodes are judiciously placed at chosen locations. In this paper, we propose a hierarchical distributed algorithm for detecting node replication attacks using a Bloom filter mechanism and a cluster head selection, we also introduce the adequate network replies in case of detection. We also present simulation results for random topologies and show that our algorithm is able to detect replication attacks in all cases.

This paper describes an automatic wireless LAN access points planning approach based on a multicriteria modelling and solving. A realistic and efficient wLAN planning approach can not only assess usual objectives based on radio coverage. It has to further implement a Quality of Service (QoS) constraint. In this work, a QoS criterion is defined as the mean available bandwidth per user, derived from a Markov-based performance evaluation model of the medium access control (MAC) layer behavior. Optimizing both coverage and QoS objectives results in finding the optimal trade-off between these concurrent criteria. Rather than optimizing a weighted sum of the optimization criteria that provides a single solution, it is herein proposed to develop a dedicated multicriteria search algorithm. Its main feature is to provide several solutions, each one representing a peculiar tradeoff between the objectives. The planning process estimates for each solution the optimal number of APs and their placement. The AP's coverage is estimated with a powerful multi-resolution radio propagation simulator previously described. This paper presents results for a QoS oriented planning process providing a predefined minimum per-user throughput. The example provided herein applies for 200 users distributed over a 12600 m2building floor.

The increase of potential applications of wireless networks in indoor environments calls for robust and efficient tools for predicting the wave propagation in such complex environments. The ParFlow technique is a discrete approach which accurately reflects the behavior of wave propagation but in turn requires very high computation and time resources. Initially this method has been implemented in a parallel architecture to reduce the computation time. A new resolution scheme is proposed to solve the discrete ParFlow equations. This approach is based on a multi-resolution formulation in the Fourier domain. The major result of our approach is that the computation complexity may be settled in two parts. Firstly, the multi-resolution pyramid is built (taking into account the geometry arrangement of the indoor environment). Secondly, the propagation from a point source is evaluated through the soft pyramidal architecture with a high gain in terms of computational efficiency. We guess that this approach could increase the interest of the ParFlow technique which could be more accurate than ray-tracing methods at least in a theoretical point of view.

The main result of this paper shows that the block-based digital medial axis transform can be computed in parallel by a constant number of calls to scan (parallel prefix) operations. This gives time- and/or work-optimal parallel implementations for the distance-based and the block-based medial axis transform in a wide variety of parallel architectures. Since only eight scan operations plus a dozen local operations are performed, the algorithm is very easy to program and use. The originality of our approach is the use of the notion of a derived grid and the oversampling of the image in order to reduce the computation of the block-based medial axis transform in the original grid to the much easier task of computing the distance based medial axis transform of the oversampling of the image on the derived grid.

Over recent years clusters of workstations have begun to replace systems of mainframes. These workstations can be used to behave as a parallel machine, and then could supply supercomputing performance at modest cost. We report an experiment of such a network within PVM framework for an industrial image analysis application. The goal of the application is to build an automated process of textile color pattern analysis. We plan to build several tools for parallel image analysis, and among those tools, we present here first the parallelization of a quantization algorithm and a way to parallelize a color image pyramid using the Parallel Virtual Machine environment. One important point is the use of the distributed storage that represents a set of workstations. This distribution allows the complete parallelization of the Read/Write operations.

The main result of this paper shows that the block-based digital medial axis transform can be computed in parallel by a constant number of calls to scan (parallel prefix) operations. This gives time- and/or work-optimal parallel algorithms for the distance-based and the block-based medial axis transform in a wide variety of parallel architectures. In order to design such an algorithm, we first demonstrate that computing a block-based medial axis transform of a binary image reduces to computing the distance based medial axis transform of an oversampling of the image, by proving that their labelings are equivalent.

Thinning is a standard kernel in pattern recognition. Both serial and parallel algorithms have been investigated. Two classes of thinning algorithms can be defined: thinning techniques that use a raster-scanning of the whole image and a mask applied on each pixel, and thinning techniques that use a contour generation of the whole image as a basic operation of the thinning operation. The goal of this paper is to increase the speed of mask thinning algorithms by increasing the size of the thinning window.

Thinning algorithms are widely used in pattern recognition. This processing stage is implemented on serial machines as well as on parallel machines. We can find two classes of thinning algorithms in the literature. The first class is composed of strongly sequential algorithms, based on contour tracing of the objects to be thinned. The second class is made of iterative processes in which fully parallel mask-based operators are used at each iteration. The weakness of the contour tracing based methods is its strong dependence upon the original picture. Particularly, in the case of very complex pictures, this algorithm becomes inefficient. The second thinning class of algorithms uses strongly regular operating processes. The general speed of those methods also depends upon the shapes to be thinned. The iteration number necessary to complete these iterative processes is proportional to the maximum thickness of the objects in the picture. The complexity of the contour of the objects does not influence the efficiency of the algorithm. The goal of this paper is to keep the regularity of the second class of thinning processes, but to reduce the amount of standard parallel-operations. We use a partition of the original picture and we dynamically reduce the range of the full-parallel operator in its sub-picture.