Localization is one of the most important topics in Wireless Sensor Networks (WSNs). This is because many fundamental techniques in WSNs (such as geographical routing, geographic key distribution, and location-based authentication) require the positions of unknown nodes. Similarly, the positions of unknown nodes play a critical role in many WSNs applications, such as monitoring applications (including environmental and health monitoring) and tracking applications (e.g., tracking objects, animals, humans, and vehicles). When a WSN is deployed in unattended and/or hostile environments, it is vulnerable to threats and risks. Many attacks exist (e.g., wormhole, sinkhole and sybil attacks) to make the estimated positions incorrect. For some applications such as military ones (e.g., battlefield surveillance) and environmental ones (e.g., forest fire detection), incorrect positions may lead to severe consequences (e.g., wrong military decisions on the battlefield, false alarms to people). Thus, we must aim to obtain a secure localization scheme to ensure the accuracy of locations. Currently, we are studying the impact of sensor collusion on existing secure localization schemes. Very little work has been done on investigating how colluding sensors can change the behavior of known attacks or even create new attacks. Depending on the nature of each secure localization algorithm, sensors could collaboratively elude the location-unknown sensors by:
a. announcing false information jointly or
b. forming an illegal position (physical) pattern (e.g. more than two sensors are collinear).
Either one of these two scenarios could lead to an erroneous calculation of a position. We start from studying a few secure localization schemes that assume no sensor collusion. We first study the possible behaviors of colluding sensors and their impact on specific secure localization schemes. We then modify the algorithms to make them sensor collusion resistant. In the next few months, we aim at categorizing all identified colluding sensor behaviors that are common to most localization schemes. As our final goal, we would like to suggest guidelines for the design of sensor collusion resistant secure localization schemes.
Sensor Deployment in Complex Region of Interest
Sensor deployment is one of the fundamental problems in WSNs. In this problem, the goal is to deploy sensors in the environment in such a way that the entire Region of Interest (ROI) is fully covered without any sensing holes. Sensing holes are the uncovered areas of a ROI that are not under the sensing range of any sensor. Using mobile sensors self deployment to cover the hostile environment is rather costly, due to the large amount of expensive mobile sensors required. In order to enable a lower cost in hostile environments, deployment strategies involving humans or mobile sensors must be avoided. Consequently, using robots deploying static sensors (which are much cheaper than mobile sensors) becomes a popular alternative. Existing research focuses on solving the problem on a ROI with boundaries forming simple shapes, such as a square or a rectangle. And the full coverage around the boundaries of ROI is often ignored. Currently, we are studying the sensor deployment problem under the following setup: a single robot deploying static sensors on a complex orthogonal ROI without requiring a GPS. The solution we aim for must:
a. guarantee the full coverage of the entire ROI;
b. minimize the total number of sensors used;
c. minimize the number of moves of the robot in order to reduce the energy consumption;
d. minimize the memory required on both the robot and the static sensors.
Cloud computing is a rising infrastructure that provides computation, software, data access, and storage resources without requiring cloud users to know the location and other details of this computing infrastructure. Cloud hosting allows customers to build scalable clouds infrastructure in multiple data centers with total control through automation and self-service. Usually networked dedicated servers, cloud servers, load balancers and cloud storage together provide such a computing infrastructure. If any node (e.g., a cloud server) on such network is not working properly, quality of service will be compromised. That's why Security, Reliability and Resiliency are the three major concerns of cloud providers and users.
The overall objective of this research consists in developing efficient and testable distributed algorithms that use mobile agents to solve non-trivial tasks in unsafe clouds, that is, in the presence of security threats. For example, a so-called black hole is a host whose behavior disposes of incoming agents or data upon their arrival, leaving no trace of such destruction. In our current research, we are studying clouds with multiple black holes in order to develop effective and efficient distributed algorithms using mobile agents in the presence of such harmful hosts. We also investigate distributed algorithms in the presence of mobile faults, such as intruders.
Multiplayer Online Game Players' Movement Prediction
We are interested in studying the relevance of distributed algorithms for Massively Multiplayer Online Games (MMOGs). In such games, I postulate that efficient distributed algorithms are required to address some of the communication and playability challenges inherent to such complex environments. The reality of poor internet quality (such as Internet lag and packet loss) can greatly hinder the playability of MMOGs. Providing seamless game play experience is a great challenge for the gaming industry. Currently we are investigating how a player's playing pattern can help to improve the accuracy of client side movement prediction algorithms.
Efficient Data Dissemination in WSNs with Mobile Sinks
Wireless sensor networks (WSNs) typically comprise a large number of power-constrained sensor modules that usually perform multi-hop data communication to a base station (sink). WSNs can be used for a number of applications ranging from surveillance and environment monitoring to health care and military operations. Several applications require that sensor nodes be left unattended for long periods of time due to cost implications or difficult access to the deployment area. Therefore, energy consumption is a major concern when designing protocols for WSNs. In static sink approaches, we can observe that nodes closer to the sink function as relays to numerous nodes.
This condition results in faster energy depletion of nodes in the vicinity of static sinks and leads to reduced network lifetime and uneven communication load. Conversely, a mobile sink is constantly changing its position, thereby resulting in even energy consumption among nodes and prolonged network lifetime. However, the problem shifts to how to find a path to a mobile sink, which induces frequent routing updates.
This project focuses in the development of energy-efficient data dissemination techniques for wireless sensor networks with mobile sinks.
Social Network-based Data Forwarding in Vehicular Sensor Networks
The convergence of vehicle and sensor networks offers viable means for data dissemination to distributed services such as traffic monitoring, safety warnings, infotainment, and proactive surveillance of roads and streets. However, data dissemination in vehicular sensor networks (VSNs) is a challenging task due to:
a) high mobility of vehicles;
b) frequent network topology changes;
c) high volume of data; and,
d) different data dissemination and application requirements.
For instance, most safety warning applications will require fast and reliable data delivery to nearby vehicles, which can be achieved by local broadcasts. A monitoring service may require continuous data delivery to a region or command centre, in which the use of broadcasts would not be feasible or efficient. Thus, another challenge lies in adapting on-the-fly to the best data dissemination approach depending on the different vehicle or user's interests.
Existing solutions address some of these issues individually and do not adapt to different services or user requirements. The heterogeneity of services, users and data requirements seems to be an opportunity to apply social network-based decision making when selecting forwarding vehicles. It might be possible to map mobility patterns as close as possible to real traffic behavior by exploiting social ties among drivers. Such mapping would be extremely useful in data forwarding decisions. Therefore, this project will investigate these issues under a unified framework and then develop novel hybrid adaptive techniques that take advantage of the strengths and deal with the weaknesses of existing data dissemination strategies. This project will focus in devising social network-based schemes that exploit "interactions" among vehicles (i.e., common visited places, encounters, mobility patterns) as well as drivers' social ties to improve data dissemination in VSNs.