Gulustan Dogan

MultiProTru: A kalman filtering based trust architecture for two-hop wireless sensor networks

Trust is an important factor in Wireless Sensor Networks in order to assess the believability of the produced data. Due to the limited computational power and energy resources of the wireless sensor networks, it is a challenge to maintain trust while using the energy efficiently. Previously we developed a trust enhancing one-hop architecture called ProTru. In order to make the architecture more efficient for multi-hop WSNs, we developed a two-hop version of our previous architecture called MultiProTru. In this architecture routing is done based on the trust values of the cluster heads. One other difference of MultiProTru is to find untrusted data we use Kalman filtering approach whereas in ProTru uses Random Sample Consensus algorithm. We compared the effectiveness of MultiProTru with ProTru, ProTruKa and a current trust architecture called an Efficient Distributed Trust Model.

Challenges of identifying second language English speakers in MOOCs

In this study, we aim to analyse English as a Second Language (ESL) and English as a First Language (EFL) MOOC participants’ engagements in a MOOC. We aim to find out key points which directly effect learners’ dropout and performance in MOOCs. We worked on a FutureLearn data which is provided by the University of Southampton. The course is Understanding Language: Learning and Teaching MOOC that was run between 2016-04-04 and 2016-05-02 is chosen for the analysis. According to the results, it is very challenging to identify who is a second language English speaker by using their location information. One of the important findings is that first language English speakers wrote longer comments. In order to identify strategies for ESL MOOC participants, which is one of the ultimate goal of our research, there is a need for much deeper analyses.

ProTru: a provenance-based trust architecture for wireless sensor networks

Summary Trust is an important component of wireless sensor networks for believability of the produced data, and trust history is a crucial asset in deciding trust of the data. In this paper, we show how provenance can be used for registering previous trust records and other information such as node type, data type, and node location. Our aim is to design a distributed trust-enhancing architecture using only local provenance during sensor fusion with a low communication overhead. Our network is cognitive in the sense that our system reacts automatically upon detecting low trust. Copyright

DynamicMultiProTru: An adaptive trust model for Wireless Sensor Networks

In Wireless Sensor Networks nodes can get untrusted after initial setup due to different reasons such as low energy, difficult environmental conditions, unexpected attacks, hardware defects. When nodes start sending faulty data, the network creates wrong observations. Errors in data can have severe impacts in wireless sensor networks such as leading to wrong decisions mission critical networks. In our earlier work, we designed an architecture for one-hop WSNs called ProTru[1]. In ProTru a static trust threshold was chosen at the time of deployment. In this work, we redeveloped the algorithm for multi-hop networks in a way that trust thresholds are chosen dynamically. We ran simulations to test the effectiveness of this proposed architecture DynamicProTru.

Provenance and trust as new factors for self-organization in a wireless sensor network

Trust can be an important component of wireless sensor networks for believability of the produced data and trust history is a crucial asset in deciding trust of the data. In our previous work, we developed an architecture called ProTru and we showed how provenance can be used for registering previous trust records and other information such as node type, data type, node location, average of historical data. We designed a distributed trust enhancing architecture using only local provenance during sensor fusion with a low communication overhead. Our network is cognitive in the sense that our system reacts automatically upon detecting low trust and restructures itself. In this work, we are extending our previous architecture by storing dataflow provenance graphs. This feature will enhance the cognitive abilities of our system by giving the network the capability of remembering past network snapshots.

Who Is the English as a Second Language Speaker in this MOOC

Massive Open Online Courses have been widely used all over the world in recent years in entirely online learning context or as blended learning on campus. Most of these courses are offered in English. A high percentage of the users, however, are speaking English as a second language. Some of the authors of this paper who either used MOOCs for blended learning or a research subject are English as a second language speakers as well. They have observed whilst teaching students at university during blended teaching using MOOCs that the students struggle in courses offered in English. This has motivated us to explore this issue in MOOCs to contribute to the pedagogy of MOOCs. The main question that we consider is how can these platforms give a better experience to second language English speakers. There are many sub-problems of this big research question. In this paper we would like to briefly present our initial findings and give an overview of the research on this area.

Uncertainty Modeling in Wireless Sensor Networks

In wireless sensor networks, the certainty of the data created by a node can change due to many reasons such as drained energy, outer factors. In addition dynamic physical environment and hardware failures (broken sensor etc.) might cause sensors to produce erred or incomplete data. Therefore raw wireless sensor network data reflects an approximate observation of the monitored environment and it can be faulty or partially wrong. Wireless sensor network applications that are used in military and health systems are used in critical decision-making. To make a healthy decision and to analyze the data correctly, the end users have to know the uncertainty level of the collected data. As a result, a model that can represent the uncertainty associated with different types of sensor data in varying models is needed. There has been some efforts of modeling uncertainty in database community. We aim to develop an uncertaintymodeling framework to aid decision-making. The methodology that we will use in this work is as follows (1) developing an uncertainty model by referring to database literature (2) designing the algorithms and equations (3) modeling finite state machines to characterize the behavior of the nodes and the transitions between different uncertainty states (4) improving a distributed architecture that defines the actions taken by the network based on the uncertainty levels (such as omitting a node with a low uncertainty level and replacing it) (5) calculating the energy and memory requirements of our architecture by carrying simulations. We have done previous work on trust computation of wireless sensor networks. In this position paper, we give a general overview of the planned research.

Investigation of social contributions of language oriented MOOC learner groups

Massive Open Online Courses (MOOCs) have attracted millions of people who are geographically dispersed. MOOCs are mainly authored in English. However, a big proportion of the participants speaks English as a foreign language. There are studies reporting that some participants struggle with understanding the language in the video lectures and are reluctant to communicate with other fellow learners. Personalisation services have been proposed to help learners who are having various difficulties in MOOCs. In order to provide help to English as a second language speakers (ESL) in MOOCs, there is a need for identifying them in case their language information is not provided. The aim of our research is to i) identify ESL participants, ii) investigate the behaviours of ESL participants, and iii) predict their performance based on their previous activities in the course to enable a personalised intervention. In our previous studies, MOOC learners were grouped based on their location and their comments. This paper investigates the overall social engagement of participants in a MOOC.

Root cause diagnosis in error-propagating networks

Various types of errors can propagate in networks, and they are usually hard to diagnose. For example, social networks spread rumors as well as useful information. Computer networks can spread Internet worms or malicious packets. In many cases, it is very hard to find the root cause (a.k.a. initial rumor spreader) of such errors without complete knowledge of the error propagation. We aim to find the root cause node when there is limited information about error propagation. We assume that there are very small number of monitor nodes in the network reporting whether error reached them or not. With this assumption, we first propose an algorithm that finds the most probable root cause node. Second, to improve the accuracy of root cause analysis, we propose another algorithm that makes use of timestamp of error reception. Finally, we study how to select monitors effectively so that root cause analysis can be accurate. With real networks from various domains, our algorithms are shown to be very effective. Copyright