I am an Assistant Professor in the CS department at the Virginia Commonwealth University. My responsibilities are teaching, research, service, and mentoring students. My research interests include mobile and pervasive computing, cyber-physical systems, social networks, wireless charging, data science, cyber security, privacy and trust, cloud, Internet of Things (IoT), crowd-sensing, network data mining and supervised learning,
Industry specialties: Mobile Packet Core (3G, LTE) gateways (GGSN, PGW, SGW, MME, SGSN, SAEGW, HA, IPSG), Femtocell gateways (HNB-GW, HeNB-GW), IMS, WiFi Offloading (ePDG, SaMOG, eWAG), Network protocols and models. Real traffic analysis and modeling. Cisco ASR 5000, ASR 5500, QvPC-SI, QvPC-DI.
Industry Expertise (4)
Writing and Editing
Areas of Expertise (4)
Mobile Computing: Wireless Networks Cyber-physical Systems Mobile Social Networks Wireless Charging
Security: Cyber Security Network Security Privacy and Trust
Mobile (3G/4G) Systems: Mobile Packet Core Cloud Mobile Data Offloading
Big Data: Internet of Things (IoT) Crowd-sensing Mobile Data Mining and Supervised Learning
Ranked with Exceptional Performance (professional)
Spotlight Paper (professional)
IEEE Student Travel Grant (professional)
Gold and Bronze Medals (professional)
In Nationwide Math Olympiad, 2008-2010.
Rensselaer Polytechnic Institute: Ph.D., Computer Science 2011
Bilkent University: M.S., Computer Engineering 2007
Bilkent University: B.S., Computer Engineering 2005
Research Grants (2)
NeTS: Small: Collaborative Research: Improving Spectrum Efficiency for Hyper-Dense IoT Networks
National Science Foundation
The emerging Internet of Things (IoT) technology will enable a whole new set of applications, imposing far reaching influence on many aspects of society, including management of health, transportation, agriculture, safety and emergency and disaster response. At the same time, the massive growth of IoT deployments poses several grand challenges to the wireless industry for their successful operation. This project aims to develop novel IoT architectures and algorithms to enable large-scale IoT deployments. In particular, through close collaboration of three academic institutions and industry, the proposed research will make advances by introducing holistic, cross-layer design framework and protocols for improving spectral/energy efficiency and latency of massive IoT networks. The research outcomes will support emerging standardization activities to enable IoT in next-generation wireless networks.
This project aims at tackling the following challenges: 1) System-Centric Waveform Design for Massive IoT: The researchers will couple stochastic geometry techniques and ambiguity functions of carrier waveforms to perform system-wide capacity analysis, compare waveform candidates for deployment suitability, develop joint equalizer and waveform designs for massive IoT deployments, and propose new waveforms having a good spectral efficiency, interference resilience, and latency characteristics. 2) Graph-based Radio Resource Management: Judicious radio resource management is critical to the proper functioning of hyper-dense IoT networks. The project will exploit weighted partitioning and matching tools from graph theory to develop waveform-aware dynamic resource allocation schemes to suppress mutual interference among a diverse set of wireless IoT links and improve spectrum utilization efficiency in single-hop and multi-hop massive IoT deployments. 3) Core network Connection Efficiency: Rethinking the core network functions for connection of massive IoT is required as they are not designed considering the IoT traffic characteristics. The project will develop and rigorously study an aggregation based connectivity model to manage multiple IoT device traffic using the same resources (i.e., bearers). For efficient groupings of the IoT devices that will share the same connection, researchers will use efficient evolutionary clustering methods, while giving preference to the interference minimizing groups obtained in earlier thrusts.
US Ignite: Collaborative Research: Focus Area 1: Rapid and Resilient Critical Data Sourcing for Public Safety and Emergency Response
National Science Foundation $380,000
In public safety and emergency response, the key for a successful recovery is on-time and reliable communication between the people reporting the incident and the emergency management authorities. Such capability should further be coupled with rapid collection of eyewitness data for successful investigation of criminals or root causes of the incident. Recently, crowd sourcing applications have proven to be promising tools to gather information with tremendous participation from the crowd. In emergency scenarios, however, such crowd sourcing solutions should be robust and resilient as the network could be congested or the underlying cellular infrastructure is damaged or temporarily lost. Moreover, an agile analysis of the massive amount of data collected should be supported with necessary equipment so that time-sensitive feedback could be provided and shared between different agencies. This project addresses these challenges by introducing a novel framework that integrates various technologies and tools for modeling and operation of large-scale crowd sourcing-based emergency response systems.
The project enhances the current systems by integrating a cloud-based rapid processing of collected data and augmenting the system by device-to-device (D2D) communications and network slicing. The project's integrated research and education plan investigates (i) large-scale critical data collection via a mobile app and management process to be used in the investigation of an emergency incident, (ii) near-real-time processing of the gathered heterogeneous data in a cloud computing environment for critical information extraction such as faces of people in the videos and photos, (iii) adoption of D2D-based communication as a complementary component to improve system resilience in case of congestion's and failures in network infrastructure, and (iv) utilization of Global Environment for Network Innovations (GENI) network slices as dedicated bandwidth for time-sensitive communication in emergency response as well as to enhance wide area resilience of the system. The project paves the way towards emergency preparedness which is a national priority; and supports progress toward smart and connected communities. The anticipated enhancements expedite the response to emergency cases, save people's lives and reduce public safety operation costs.
CMSC 628 Mobile Networks
The course will assume undergraduate-level background in algorithms, programming (e.g., Java), calculus, and probability. Upon successful completion of this course, the student will be able to understand the major concepts about mobile networks; be familiar with various mobile network applications (e.g., ad hoc and sensor networks, mobile social networks, delay tolerant networks, vehicular networks and cellular networks); learn how to model mobile networks with stochastic processes and real datasets; be able to use different networking simulators; understand various routing algorithms and analyze their behavior.
• Different mobile network applications (e.g., Ad hoc and sensor networks, Delay tolerant networks, Mobile Social Networks, Vehicular networks) and challenges
• Device-to-Device communication technologies (e.g., Bluetooth, WiFi-Direct, LTE-direct)
• Routing algorithms for content distribution and delivery
• Mobility models
• Mathematical tools to analyze and model mobile networks
• Network simulators (ns-2, ONE etc.)
• Data driven simulations and evaluation
• Emerging Networks and Technologies (Internet of Things, Machine to Machine Networks, Connected cars, DSRC, Aerial networks)
CMSC 428 Mobile Programming - iOS
This course covers the fundamentals of Swift and XCode for programming and design of iOS applications. Students develop different types of iOS applications (game, web based etc.) completely functional throughout the course. They also test the developed applications on simulator and on actual device.
• Introduction to Swift, Xcode and iOS Frameworks
• Swift basics: variables, operations, loops, if statements, functions, optionals
• Object oriented programming, polymorphism, inheritance
• Auto Layout, Stack view, Scroll view, Interface development
• ViewControllers and Segues
• Tableview, Collectionview
• Model, View, Controller (MVC)
• Touches and Gestures
• Data Persistence
• Animations and SpriteKit
• MapKit, Core Location, CoreMotion
• Web services, networking, multipeer connectivity
• Submission to Apple Store
CMSC 401 Algorithm Analysis with Advanced Data Structures
This course covers the basic principles of algorithm design and analysis. Main topics covered:
• Foundations of algorithm and complexity analysis
• Divide and conquer approach
• Data Structures
• Sorting and Order Statistics
• Advanced design and analysis techniques
• Graph algorithms
• Dynamic programming
Selected Articles (5)
Eyuphan Bulut, Mithat Kisacikoglu, and Kemal Akkaya
Direct Vehicle-to-vehicle (V2V) charge sharing system has the potential to provide more flexibility to electric vehicle (EV) charging without depending on the charging station infrastructure or building designated parking lots. It can also provide an opportunity to shift peak time utility load to off-peak times. However, the assignment between the EVs that demand energy and the EVs with surplus energy or existing charging stations is a challenging problem as it has to be performed in real-time considering their spatio-temporal distribution, availability and grid load. In this paper, we study this assignment problem specifically in a supplier non-intrusive scenario (without changing their mobility) and aim to understand the potential benefits of a direct V2V charge sharing system.
To this end, we present two new algorithms to match the demander EVs to suppliers for charging. In the first one, the maximum system benefit is targeted under multiple system objectives with different priorities and considering the grid load. In the second one, individual EV priorities are taken into account and a stable matching process depending on predefined user preferences is provided. We conduct extensive simulations using real user commuting patterns and existing charging station locations in three different cities together with a newly developed probabilistic EV charging behavior model and EV mobility. Simulation results show that direct V2V charge sharing can reduce the energy consumption of EVs by 20-35% by providing a closer charging facility compared to charging station only case, offload grid charging power in peak times by 35-55%, and help sustain up to 2.2x EV charging requests without building new charging infrastructure, while causing a marginal increase (i.e., 3-9%) in the energy cycling of supplier EVs. The results also show that the proposed matching algorithms offer 80-90% more energy consumption reduction compared to the intrusive V2V charging at designated parking lots.
Aashish Dhungana, Tomasz Arodz, and Eyuphan Bulut
In this paper, we investigate the utilization of peer-to-peer wireless energy sharing to relieve the users from the burden of cord-based charging. The devices of users can make use of energy available from other users' devices based on their meeting patterns so that the battery level of their devices could be maintained within an acceptable level without the need of charging it through a cable frequently. We first use dynamic programming-based optimization to find out the minimum number of cord-based charging sessions to obtain the highest possible mobile charging relief through collaborative charge sharing among pairs of nearby user devices. Then, we map our problem to roommate matching problem and find out the best matching among users that will achieve the highest network-wide relief while satisfying all users with their assigned partners. With an extensive empirical analysis based on real device charging patterns and meeting patterns between mobile users, we evaluate the charging relief offered to users in various scenarios. The results show that users can get up to 13-17% relief from their charging burden using cooperative energy exchanges without changing their existing usage habits.
Aashish Dhungana and Eyuphan Bulut
In mobile social networks, the mobility and connectivity of nodes are often non-deterministic. Source and destination nodes may not have a meeting opportunity and the content dissemination and delivery most of the time require the cooperation of nodes. However, this causes nodes spend energy, thus, they may be reluctant to participate in the dissemination process to conserve energy. One approach to motivate user participation is to transfer energy for their service so that their potential loss is compensated. However, this makes routing problem much challenging as the source node needs to decide not only the best relay nodes but also the amount of energy transfer to them. In this paper, we study this energy sharing based content delivery problem in mobile social networks. To this end, we assume that a node is willing to carry the content as long as the energy received for this delivery lasts, after which it drops the content (i.e., time-to-live). We utilize optimal stopping theory and dynamic programming to model the content delivery problem under this energy sharing paradigm between the nodes. The simulation results show that energy sharing based content delivery can potentially increase the routing performance under certain settings.
Fatih Yucel, Kemal Akkaya and Eyuphan Bulut
Electric Vehicle (EV) charging takes longer time and happens more frequently compared to refueling of fossil-based vehicles. This requires in-advance scheduling on charging stations depending on the route of the demander EVs for efficient resource allocation. However, such scheduling and frequent charging may leak sensitive information about the users which may expose their driving patterns, whereabouts, schedules, etc. The situation is compounded with the proliferation of EV chargers such as V2V charging where any two EVs can charge each other through a charging cable. In such cases, the matching of these EVs is typically done in a centralized manner which exposes private information to third parties which do the matching. To address this issue, in this paper, we propose an efficient and privacy-preserving distributed matching of demander EVs with charge suppliers (i.e., public/private stations, V2V chargers) using bichromatic mutual nearest neighbor (BMNN) assignments. To this end, we use partially homomorphic encryption-based BMNN computation through local communication (e.g., DSRC or LTE-direct) between users while hiding their locations. The proposed matching algorithm provides not only a satisfactory assignment for all parties but also achieves an efficient matching in dynamic environments where new demanders and suppliers show up and some leave. The simulation results indicate that the proposed matching of suppliers and demanders can be achieved in a distributed fashion within reasonable computation and convergence times while preserving privacy of users. Moreover, due to the nature of its design, it provides a more efficient matching process for dynamic environments compared to standard stable matching algorithm, reducing the average waiting time for users until matching.
Fatih Yucel, and Eyuphan Bulut
With the proliferation of mobile devices having BLE capability and the introduction of Beacon technology, crowdsourcing based approaches have recently emerged as a promising solution for localization of lost objects or individuals (e.g., children or elders). By attaching affordable Beacon tags to them, objects of care could be tracked and localized by the user devices in the proximity. While such a crowd GPS service has gained popularity recently, it did not extend beyond passive mode in which localization is achieved in the background without intruding the mobility of users.
In this paper, we study the localization of lost objects through the crowd GPS service in an active manner. We propose clustering of users in a Beacon tag network based on the benefits they can receive from each other in terms of the localization of their lost items. A new metric is developed to quantify this benefit and the users that can provide most of the total possible benefits to each other are then grouped together so that they can provide active localization service for only the users that can provide high benefit to them. The clustering of users is achieved based on both a greedy heuristic based algorithm and a genetic algorithm. Extensive simulation results are conducted utilizing both synthetic data and real location based social network datasets. The results show the effective partitioning of the users under different user counts and groups while valuing the privacy of users at its maximum by limiting the number of interactions between users.