Sahin Cem Geyik - Academia.edu (original) (raw)

Papers by Sahin Cem Geyik

Social networks analysis includes examining the actions of entities in a social setting. These ac... more Social networks analysis includes examining the actions of entities in a social setting. These actions can be either interactions between entities (e.g. talking, exchanging items etc.), or actions which do not include interactions, but nevertheless are happening in a social context, hence are influenced by social relations. Such actions often contain behavioral patterns that are specific to the actions involved. It is important to understand such patterns to be able to model social environments reliably. In this work, we introduce a novel method for modeling and classifying behavior of nodes in a social network using Probabilistic Context Free Grammars (PCFGs). Informally, PCFGs are regular context free grammars (consisting of START symbol, terminals, nonterminals

This paper investigates the application of model-driven techniques to the construction and compos... more This paper investigates the application of model-driven techniques to the construction and composition of services on sensor networks. We present a model that gives the user a visual representation of a service, that can be annotated with semantic information (for example performance characteristics, deployment constraints, policies and rules, etc.) using an appropriate extensible user-oriented vocabulary. We propose the use of UML 2.0 Activity Diagrams as our graphical notation, with semantic annotations represented as properties. We show the transformation of the UML model to a semantic representation conforming to an appropriate ontology and use this as the core model for subsequent static and dynamic analysis. We show how the core model can be used to generate domain-specific representations suitable for input to analysis and development tools. Two examples are given: (i) generation of a Performance Evaluation Process Algebra (PEPA) [1] model, and (ii) generation of a specificat...

Lecture Notes in Computer Science, 2008

This paper presents a biologically inspired routing protocol called Self Selective Routing with p... more This paper presents a biologically inspired routing protocol called Self Selective Routing with preferred path selection (SSR(v3)). Its operation resembles the behavior of a biological ant that finds a food source by following the strongest pheromone scent left by scout ants at each fork of a path. Likewise, at each hop of a multi-hop path, a packet using the Self Selective Routing (SSR) protocol moves to the node with the shortest hop distance to the destination. Each intermediate node on a route to the destination uses a transmission back-off delay to select a path to follow for each packet of a flow. Neither an ant nor a packet knows in advance the route that each will follow as it is decided at each step. Therefore, when a route becomes severed by a failure, they can dynamically and locally adjust their routing to traverse the shortest surviving path. Preferred path selection reduces transmission delay by essentially removing back-off delay for the node that carried the previous packet of the same flow. The results reported here for both simulation and execution of a MicaZ mote implementation, show that this is an efficient and fault-tolerant protocol with small transmission delay, high reliability and high delivery rate.

Workshop on Wireless Mobile Multimedia, 2010

Delay tolerant networks are characterized by the sporadic connectivity between their nodes and th... more Delay tolerant networks are characterized by the sporadic connectivity between their nodes and therefore the lack of stable end-to-end paths from source to destination. Since the future node connections are mostly unknown in these networks, opportunistic forwarding is used to deliver messages. However, making effective forwarding decisions using only the network characteristics (i.e. average intermeeting time between nodes) extracted from

In the domain of online advertising, our aim is to serve the best ad to a user who visits a certa... more In the domain of online advertising, our aim is to serve the best ad to a user who visits a certain webpage, to maximize the chance of a desired action to be performed by this user after seeing the ad. While it is possible to generate a different prediction model for each user to tell if he/she will act on a given ad, the prediction result typically will be quite unreliable with huge variance, since the desired actions are extremely sparse, and the set of users is huge (hundreds of millions) and extremely volatile, i.e., a lot of new users are introduced everyday, or are no longer valid. In this paper we aim to improve the accuracy in finding users who will perform the desired action, by assigning each user to a cluster, where the number of clusters is much smaller than the number of users (in the order of hundreds). Each user will fall into the same cluster with another user if their event history are similar. For this purpose, we modify the probabilistic latent semantic analysis (pLSA) model by assuming the independence of the user and the cluster id, given the history of events. This assumption helps us to identify a cluster of a new user without re-clustering all the users. We present the details of the algorithm we employed as well as the distributed implementation on Hadoop, and some initial results on the clusters that were generated by the algorithm.

Ground/Air Multisensor Interoperability, Integration, and Networking for Persistent ISR II, 2011

Our previous work has explored the application of enterprise middleware techniques at the edge of... more Our previous work has explored the application of enterprise middleware techniques at the edge of the network to address the challenges of delivering complex sensor network solutions over heterogeneous communications infrastructures. In this paper, we develop this approach further into a practicable, semantically rich, model-based design and analysis approach that considers the sensor network and its contained services as a service-oriented architecture. The proposed model enables a systematic approach to service composition, analysis (using domain-specific techniques), and deployment. It also enables cross intelligence domain integration to simplify intelligence gathering, allowing users to express queries in structured natural language (Controlled English).

2012 IEEE Ninth International Conference on Services Computing, 2012

Sensor applications are typically composed of a number of functional components that run distribu... more Sensor applications are typically composed of a number of functional components that run distributedly on the nodes of a sensor network, communicating and interacting with one another. Service composition is emerging as a viable approach towards the automatic synthesis of such sensor applications. However, for service composition to be practical, it has to comply with policies that define security and management constraints on the use of these service components and the interconnections amongst them. Prior research efforts have primarily focused on efficient evaluation of security policies during the composition process, which is not sufficient when generic network management constraints need to be expressed and evaluated. In this work, we propose a policy model and evaluation approach that enables us to define and check attribute-based policies, for controlling the sensor service composition process. Attributebased policies are generic and allows us to express a wider spectrum of constraints than currently possible. Using this model and based on a previously-proposed sensor service composition algorithm, we introduce a policy evaluation method that allows for efficient checking of policy constraints. We further present a novel implementation of the proposed approach in the IBM Sensor Fabric, a middleware framework that simplifies the development of distributed, sensor network services. We also present preliminary performance evaluation results using our prototype.

The 7th IEEE International Conference on Mobile Ad-hoc and Sensor Systems (IEEE MASS 2010), 2010

In a delay tolerant network (DTN), nodes are connected intermittently and the future node connect... more In a delay tolerant network (DTN), nodes are connected intermittently and the future node connections are mostly unknown. Since in these networks, a fully connected path from source to destination is unlikely to exist, message delivery relies on opportunistic routing. However, effective forwarding based on a limited knowledge of contact behavior of nodes is challenging. Most of the previous studies looked at only the pairwise node relations to decide routing. In contrast, in this paper, we analyze the correlation between the meetings of each node with other nodes and focus on the utilization of this correlation for efficient routing of messages. We introduce a new metric called conditional intermeeting time, which computes the average intermeeting time between two nodes relative to a meeting with a third node using only the local knowledge of the past contacts. Then, we show how we can utilize the proposed metric on the existing DTN routing protocols to improve their performance. For shortest-path based routing protocols in DTNs, we propose to route messages over conditional shortest paths in which the link cost between nodes are defined by conditional intermeeting times. Moreover, for metric-based forwarding protocols, we propose to use conditional intermeeting time as an additional delivery metric while making forwarding decisions of messages. Our trace-driven simulations on three different datasets show that the modified algorithms perform better than the original ones.

Pervasive and Mobile Computing, 2012

This paper describes the use of market mechanisms for resource allocation in pervasive sensor app... more This paper describes the use of market mechanisms for resource allocation in pervasive sensor applications to maximize their Value of Information (VoI), which combines the objectively measured Quality of Information (QoI) with the subjective value assigned to it by the users. The unique challenge of pervasive sensor applications that we address is the need for adjusting resource allocation in response to the changing application requirements and evolving sensor network conditions. We use two market mechanisms: auctions at ...

Proceedings of 20th ACM SIGKDD Conference on Knowledge Discovery and Data Mining - ADKDD'14, 2014

Budget allocation in online advertising deals with distributing the campaign (insertion order) le... more Budget allocation in online advertising deals with distributing the campaign (insertion order) level budgets to different sub-campaigns which employ different targeting criteria and may perform differently in terms of return-on-investment (ROI). In this paper, we present the efforts at Turn on how to best allocate campaign budget so that the advertiser or campaign-level ROI is maximized. To do this, it is crucial to be able to correctly determine the performance of sub-campaigns. This determination is highly related to the action-attribution problem, i.e. to be able to find out the set of ads, and hence the sub-campaigns that provided them to a user, that an action should be attributed to. For this purpose, we employ both last-touch (last ad gets all credit) and multi-touch (many ads share the credit) attribution methodologies. We present the algorithms deployed at Turn for the attribution problem, as well as their parallel implementation on the large advertiser performance datasets. We conclude the paper with our empirical comparison of last-touch and multi-touch attribution-based budget allocation in a real online advertising setting.

2011 IEEE International Conference on Pervasive Computing and Communications Workshops (PERCOM Workshops), 2011

This paper examines the possible uses of different market mechanisms for resource allocation at d... more This paper examines the possible uses of different market mechanisms for resource allocation at different levels of Wireless Sensor Network (WSN) architecture. The goal is to maximize the Value of Information (VoI) for WSN users. We discuss three different levels of WSN architecture. The lowest level focuses on individual nodes and their basic functions of sensing and routing. We give an example showing how the use of auctions at individual nodes can help to efficiently perform these functions. The middle level focuses on services that are abstractions of applications running on sensors. Complex applications are composed automatically from basic ones. We discuss the use of switch options to address some of the challenges arising in such dynamic service composition. Finally, we consider the highest level -network deployment and sharing -and conjecture that options may be valuable in creating proper incentives for rational deployment and sharing of WSNs.

2011 IEEE International Conference on Services Computing, 2011

Service-oriented Architecture (SOA) for sensor network applications aims at providing composable ... more Service-oriented Architecture (SOA) for sensor network applications aims at providing composable sensor network services supporting functionality within a specific application domain together with tools for service composition, so more complex functionalities can be composed of component services. In a distributed environment, such a scheme works by having a given component service choose other components that provide the data that it needs to perform its service. In this paper, we propose to use real options theory for selecting component services. Real options are designed to reduce the risk associated with an investment by delaying the investment decision for a certain period of time or by allowing for the substitutions of initial investment. Thus, they enhance managerial flexibility and add to the overall value of a project but at the same time they incur certain costs. It is natural to think about activated services as investments, and we apply the switch options subset of the real options methodology to manage the risks of high cost that may result from the low reliability of sensors and sensor networks. Furthermore, we compare our approach with several service selection methods and show the advantage of the option-based methodology.

IEEE INFOCOM 2009 - The 28th Conference on Computer Communications, 2009

Modern military and civilian surveillance applications should provide end users with the high lev... more Modern military and civilian surveillance applications should provide end users with the high level representation of events observed by sensors rather than with the raw data measurements. Hence, there is a need for a system that can infer higher level meaning from collected sensor data. We demonstrate that probabilistic context free grammars (PCFGs) can be used as a basis for such a system. To recognize events from raw sensor network measurements, we use a PCFG inference method based on Stolcke(1994) and Chen(1996). We present a fast algorithm for deriving a concise probabilistic context free grammar from the given observational data. The algorithm uses an evaluation metric based on Bayesian formula for maximizing grammar a posteriori probability given the training data. We also present a real-world scenario of monitoring a parking lot and the simulation based on this scenario. We described the use of PCFGs to recognize events in the results of such a simulation. We finally demonstrate the deployment details of such an event recognition system.

MILCOM 2008 - 2008 IEEE Military Communications Conference, 2008

Diverse security applications often require monitoring of a narrow passage, such as an indoor cor... more Diverse security applications often require monitoring of a narrow passage, such as an indoor corridor, a tunnel, a bridge; either to protect critical assets at the end of such a passage or to control the passage over it, or both. Often, sensors are arranged in a vector along such a passage and are capable of registering the crossing of a target but not its identity. In this paper, we consider coordinated tracking by such a vector of sensors where each sensor learns timings of objects passing at monitored spots from its predecessors. Hence, the problem we are solving could be formulated as an assignment or matching of target identities to arrival times of the targets at the subsequent sensors in a vector.

2010 IEEE International Conference on Services Computing, 2010

Service modeling and composition is a fundamental method for offering advanced functionality by c... more Service modeling and composition is a fundamental method for offering advanced functionality by combining a set of primitive services provided by the system. Unlike in the case of web services for which there is an abundance of reliable resources, in sensor networks, the resources are constrained and communication among nodes is error-prone and unreliable. Such a dynamic environment requires a continuous adaptation of the composition of services. In this paper, we first propose a graph-based model of sensor services that maps to the operational model of sensor networks and is amenable to analysis. Based on this model, we formulate the process of sensor service composition as a cost-optimization problem, which is NP-complete. We then propose two heuristic methods for its solution, the top-down and the bottom-up, and discuss their centralized and distributed implementations. Using simulations, we evaluate their performance.

2012 IEEE Ninth International Conference on Services Computing, 2012

Service composition in sensor networks combines elementary services with a specific functionality... more Service composition in sensor networks combines elementary services with a specific functionality to create a service with higher level functionality. The previous efforts in automating composition were sending full information about all services across the entire sensor network, creating a security risk and imposing significant communication overhead. Furthermore, learning based composition or error detection methods do not consider global information, leading to inefficiencies in the generated composition graphs. In this paper, we propose a probabilistic context-free grammar (PCFG) based modeling technique to construct service compositions. The successful compositions created for the given application are treated as statements belonging to an efficient composition PCFG of this application. The given set of such compositions is used to derive this PCFG automatically. Future composition could be then easily constructed with the help of such PCFG. We present our methodology for achieving such modeling and provide examples of its use to demonstrate its advantage over previous work. We also evaluate the resulting improvements in performance of compositions and in the costs of their creation.

2010 13th International Conference on Information Fusion, 2010

Identifying the behavioral patterns in a social network setting is beneficial to understand how p... more Identifying the behavioral patterns in a social network setting is beneficial to understand how people behave in certain application domains. Such patterns can also be utilized to characterize social signals such as social roles from interactions. In this work, we examine how probabilistic context free grammars (PCFGs) can be utilized to model interactions and role taking in a social network. We describe how to automatically build a PCFG given a set of interactions as the training data. Our experiments on the Mission Survival Corpus 1 (MSC-1) dataset show that PCFGs are a concise way of modeling social entity behaviors and are useful in understanding the probability distribution of interactions as well as the behavior types that are observed.

Pervasive and Mobile Computing, 2013

ABSTRACT In a delay tolerant network (DTN), nodes are connected intermittently and the future nod... more ABSTRACT In a delay tolerant network (DTN), nodes are connected intermittently and the future node connections are mostly not known. Therefore, effective forwarding based on limited knowledge of contact behavior of nodes is challenging. Most of the previous studies assumed that mobility of a node is independent from mobility of other nodes and looked at only the pairwise node relations to decide routing. In contrast, in this paper, we analyze the temporal correlation between the meetings of each node with other nodes and utilize this correlation for efficient routing. We introduce a new metric called conditional intermeeting time (CIT), which computes the average intermeeting time between two nodes relative to a meeting with a third node. Then, we modify existing DTN routing protocols using the proposed metric to improve their performance. Extensive simulations based on real and synthetic DTN traces show that the modified algorithms perform better than the original ones.

IEEE Transactions on Services Computing, 2000

... in Sensor Networks Sahin Cem Geyik, Student Member, IEEE, Boleslaw K. Szymanski, Fellow, IEEE... more ... in Sensor Networks Sahin Cem Geyik, Student Member, IEEE, Boleslaw K. Szymanski, Fellow, IEEE, and Petros Zerfos, Member, IEEE, ... E-mail: {geyiks,szymansk} @cs.rpi.edu • Petros Zerfos is with IBM TJ Watson Research Center, Hawthorne, NY, 10532. ...

IEEE Transactions on Mobile Computing, 2000

Modeling of the mobility patterns arising in computer networks requires a compact and faithful re... more Modeling of the mobility patterns arising in computer networks requires a compact and faithful representation of the mobility data collected from observations and measurements of the relevant network applications. This data can range from the information on the mobility of the agents that are being monitored by a wireless network to mobility information of nodes in mobile network applications. In this paper, we examine the use of probabilistic context free grammars as the modeling framework for such data. We present a fast algorithm for deriving a concise probabilistic context free grammar from the given training data. The algorithm uses an evaluation metric based on Bayesian formula for maximizing grammar a posteriori probability given the training data. We describe the application of this algorithm in two mobility modeling domains: (i) recognizing mobility patterns of monitored agents in different event datasets collected by sensor networks, and (ii) modeling and generating node movements in mobile networks. We also discuss the model's performance in simulations utilizing both synthetic and real world mobility traces.