ieee projects 2009

8/3/2019 IEEE Projects 2009

1/32

IEEE PROJECT TITLES

IEEE TITLES CONTENT

NETWORKING ....................................................................... 3RESEQUENCING ANALYSIS OF STOP-AND-WAIT ARQ FOR PARALLEL MULTICHANNELCOMMUNICATIONS ..................................................................................................................................3RESOURCE ALLOCATION IN OFDMA WIRELESS COMMUNICATIONS SYSTEMS SUPPORTINGMULTIMEDIA SERVICES ............................................................................................................................3ANALYSIS OF SHORTEST PATH ROUTING FOR LARGE MULTI-HOP WIRELESS NETWORKS .....................4SECURE AND POLICY-COMPLIANT SOURCE ROUTING .............................................................................4NODE ISOLATION MODEL AND AGE-BASED NEIGHBOR SELECTION IN UNSTRUCTURED P2PNETWORKS .................................................................................................................................. ...... ......4MOBILITY MANAGEMENT APPROACHES FOR MOBILE IP NETWORKS: PERFORMANCE COMPARISONAND USE RECOMMENDATIONS ................................................................................................................4SINGLE-LINK FAILURE DETECTION IN ALL-OPTICAL NETWORKS USING MONITORING CYCLES ANDPATHS ......................................................................................................................................................5MULTIPLE ROUTING CONFIGURATIONS FOR FAST IP NETWORK RECOVERY ...........................................5VIRUS SPREAD IN NETWORKS .................................................................................................................5PRESTO: FEEDBACKDRIVEN DATA MANAGEMENT IN SENSOR NETWORKS ............................................5EXPLICIT LOAD BALANCING TECHNIQUE FOR NGEO SATELLITE IP NETWORKS WITH ON-BOARDPROCESSING CAPABILITIES .....................................................................................................................6DELAY ANALYSIS FOR MAXIMAL SCHEDULING WITH FLOW CONTROL IN WIRELESS NETWORKS WITHBURSTY TRAFFIC ......................................................................................................................................6MINING FILE DOWNLOADING TIME IN STOCHASTIC PEER TO PEER NETWORKS .....................................6ENHANCING SEARCH PERFORMANCE IN UNSTRUCTURED P2P NETWORKS BASED ON USERS'COMMON INTEREST .................................................................................................................................7BRA: A BIDIRECTIONAL ROUTING ABSTRACTION FOR ASYMMETRIC MOBILE AD HOC NETWORKS ........7RATE & DELAY GUARANTEES PROVIDED BY CLOSE PACKET SWITCHES WITH LOAD BALANCING ....... ..7GEOMETRIC APPROACH TO IMPROVING ACTIVE PACKET LOSS MEASUREMENT .....................................7PERFORMANCE OF A SPECULATIVE TRANSMISSION SCHEME FOR SCHEDULING LATENCY REDUCTION 8RATE ALLOCATION & NETWORK LIFETIME PROBLEM FOR WIRELESS SENSOR NETWORKS ................. ..8STATISTICAL TECHNIQUES FOR DETECTING TRAFFIC ANOMALIES THROUGH PACKET HEADER DATA . .8EFFICIENT ROUTING IN INTERMITTENTLY CONNECTED MOBILE NETWORKS: THE MULTIPLE COPY CASE................................................................................................................................................................9

TWO TECHNIQUES FOR FAST COMPUTATION OF CONSTRAINED SHORTEST PATHS ..............................9

PROBABILISTIC PACKET MARKING FOR LARGE-SCALE IP TRACE BACK ...................................................9DUAL-LINK FAILURE RESILIENCY THROUGH BACKUP LINK MUTUAL EXCLUSION ..................................10A DISTRIBUTED DATABASE ARCHITECTURE FOR GLOBAL ROAMING IN NEXT-GENERATION MOBILENETWORKS ............................................................................................................................................10NETWORK BORDER PATROL: PREVENTING CONGESTION COLLAPSE AND PROMOTING FAIRNESS IN

THE INTERNET ............................................................................................................................ ...... .....10

MOBILE COMPUTING ............................................................ 11A TABU SEARCH ALGORITHM FOR CLUSTER BUILDING IN WIRELESS SENSOR NETWORKS .................11ROUTE STABILITY IN MANETS UNDER THE RANDOM DIRECTION MOBILITY MODEL .............................11GREEDY ROUTING WITH ANTI-VOID TRAVERSAL FOR WIRELESS SENSOR NETWORKS ........................11CELL BREATHING TECHNIQUES FOR LOAD BALANCING IN WIRELESS LANS ........................................12CAR: CONTEXT-AWARE ADAPTIVE ROUTING FOR DELAY-TOLERANT MOBILE NETWORKS ...................12LOCAL CONSTRUCTION OF NEAR-OPTIMAL POWER SPANNERS FOR WIRELESS AD-HOC NETWORKS . 12BIASED RANDOM WALKS IN UNIFORM WIRELESS NETWORKS ..............................................................12INFORMATION CONTENT-BASED SENSOR SELECTION AND TRANSMISSION POWER ADJUSTMENT FORCOLLABORATIVE TARGET TRACKING ....................................................................................................13

ENERGY MAPS FOR MOBILE WIRELESS NETWORKS:COHERENCE TIME VERSUS SPREADING PERIOD . 13RANDOMCAST: AN ENERGY EFFICIENT COMMUNICATION SCHEME FOR MOBILE AD HOC NETWORKS 13EFFICIENT RESOURCE ALLOCATION FOR WIRELESS MULTICAST ..........................................................13INTRUSION DETECTION IN HOMOGENEOUS & HETEROGENEOUS WIRELESS SENSOR NETWORKS .. .. .14LOCATION BASED SPATIAL QUERY PROCESSING IN WIRELESS BROADCAST ENVIRONMENTS .............14BANDWIDTH ESTIMATION FOR IEEE 802.11 BASED ADHOC NETWORK ................................................14BENEFIT-BASED DATA CACHING IN AD HOC NETWORKS ......................................................................14LOCALIZED SENSOR AREA COVERAGE WITH LOW COMMUNICATION OVERHEAD ...............................14EFFICIENT RESOURCE ALLOCATION FOR WIRELESS MULTICAST ..........................................................15AN ACKNOWLEDGMENT-BASED APPROACH FOR THE DETECTION OF ROUTING MISBEHAVIOR INMANETS .............................................................................................................................................. ...15ROUTE RESERVATION IN AD HOC WIRELESS NETWORKS .....................................................................15


2/32

DISTRIBUTED CACHE UPDATING FOR THE DYNAMIC SOURCE ROUTING PROTOCOL ...........................16

PARALLEL AND DISTRIBUTED SYSTEMS ................................. 16DYNAMIC SEARCH ALGORITHM IN UNSTRUCTURED PEER-TO-PEER NETWORKS ..................................16FLEXIBLE DETERMINISTIC PACKET MARKING: AN IP TRACEBACK SYSTEM TO FIND THE REAL SOURCEOF ATTACKS ..........................................................................................................................................16DISTRIBUTED ALGORITHMS FOR CONSTRUCTING APPROXIMATE MINIMUM SPANNING TREES INWIRELESS SENSOR NETWORKS .............................................................................................................17A FAITHFUL DISTRIBUTED MECHANISM FOR SHARING THE COST OF MULTICAST TRANSMISSIONS .. ..17DYNAMIC ROUTING WITH SECURITY CONSIDERATIONS ........................................................................17COMPACTION OF SCHEDULES AND A TWO-STAGE APPROACH FOR DUPLICATION-BASED DAGSCHEDULING ..................................................................................................................................... ....18DETECTING MALICIOUS PACKET LOSSES ..............................................................................................18QUIVER: CONSISTENT OBJECT SHARING FOR EDGE SERVICES .............................................................18HBA DISTRIBUTED METADATA MANAGEMENT FOR LARGE SCALE CLUSTER BASED STORAGE SYSTEM..............................................................................................................................................................19DYNAMIC LOAD BALANCING IN DISTRIBUTED SYSTEMS IN THE PRESENCE OF DELAYS: AREGENERATION-THEORY APPROACH ....................................................................................................19OCGRR: A NEW SCHEDULING ALGORITHM FOR DIFFERENTIATED SERVICES NETWORKS ................ ...19PFUSION: A P2P ARCHITECTURE FOR INTERNET-SCALE CONTENT-BASED SEARCH AND RETRIEVAL . .19

DEPENDABLE AND SECURE COMPUTING ................................ 19THE EFFECTIVENESS OF CHECKSUMS FOR EMBEDDED NETWORKS ....................................................20A PRECISE TERMINATION CONDITION OF THE PROBALASTIC PACKET MARKING ALGORITHM .......... ...20MODELING & AUTOMATED CONTAINMENT OF WORMS ........................................................................20

TRUST WORTHY COMUTING UNDER RESOURCE CONSTRAINTS WITH THE DOWN POLICY ..................20TEMPORAL PORTIONING OF COMMUNICATION RESOURCES IN AN INTEGRATED ARCHITECTURE ..... ..21CONSTRUCTING INTER-DOMAIN PACKET FILTERS TO CONTROL IP SPOOFING BASED ON BGP UPDATES..............................................................................................................................................................21AN ADAPTIVE PROGRAMMING MODEL FOR FAULT-TOLERANT DISTRIBUTED COMPUTING ................ ..21HYBRID INTRUSION DETECTION WITH WEIGHTED SIGNATURE GENERATION OVER ANOMALOUSINTERNET EPISODES(HIDS) ...................................................................................................................21

KNOWLEDGE AND DATA ENGINEERING/ DATA MINING ............ 22AN EFFICIENT CLUSTERING SCHEME TO EXPLOIT HIERARCHICAL DATA IN NETWORK TRAFFICANALYSIS ...............................................................................................................................................22PROTECTION OF DATABASE SECURITY VIA COLLABORATIVE INFERENCE DETECTION .........................22HARDWARE ENHANCED ASSOCIATION RULE MINING WITH HASHING AND PIPELINING .......................22WATERMARKING RELATIONAL DATABASES USING OPTIMIZATION-BASED TECHNIQUES .....................23

TRUTH DISCOVERY WITH MULTIPLE CONFLICTING INFORMATION PROVIDERS ON THE WEB ............. .23

SOFTWARE ENGINEERING .................................................... 23ATOMICITY ANALYSIS OF SERVICE COMPOSITION ACROSS ORGANIZATIONS ......................................23

OPTIMIZED RESOURCE ALLOCATION FOR SOFTWARE RELEASE PLANNING .........................................23AUTOMATIC EXTRACTION OF HEAP REFERENCE PROPERTIES IN OBJECT-ORIENTED PROGRAMS ........24USING THE CONCEPTUAL COHESION OF CLASSES FOR FAULT PREDICTION IN OBJECT ORIENTEDSYSTEMS ................................................................................................................................................24

THE EFFECT OF PAIRS IN PROGRAM DESIGN TASKS .............................................................................24ESTIMATION OF DEFECTS BASED ON EFECT DECAY MODEL: ED3M .....................................................24

IMAGE PROCESSING ............................................................ 25EFFICIENT 2-D GRAY SCALE MORPHOLOGICAL TRANSFORMATIONS WITH ARBITRALY FLATSTRUCTURING ELEMENTS .....................................................................................................................25VISION BASED PROCESSING FOR REAL TIME 3-D DATA ACQUISITION BASED CODE STRUCTUREDLIGHT .....................................................................................................................................................25ACTIVE LEARNING METHODS FOR INTERACTIVE IMAGE RETRIEVAL .....................................................25DIGITAL IMAGE PROCESSING TECHNIQUES FOR THE DETECTION AND REMOVAL OF CRACKS INDIGITIZED PAINTINGS ............................................................................................................................26STRUCTURE AND TEXTURE FILLING-IN OF MISSING IMAGE BLOCKS IN WIRELESS TRANSMISSION ANDCOMPRESSION APPLICATIONS ...............................................................................................................26

APPLICATION OF BPCS STEGANOGRAPHY TO WAVELET COMPRESSED VIDEO .....................................26ORTHOGONAL DATA EMBEDDING FOR BINARY IMAGES IN MORPHOLOGICAL TRANSFORM DOMAIN- AHIGH-CAPACITY APPROACH ................................................................................................................ ...27

MULTIMEDIA ....................................................................... 27A NOVEL FRAMEWORK FOR SEMANTIC ANNOTATION AND PERSONALIZED RETRIEVAL OF SPORTSVIDEO ....................................................................................................................................... ...... .......27

PATTERN ANALYSIS AND MACHINE INTELLIGENCE ................. 27FACE RECOGNITION USING LAPLACIAN FACES .....................................................................................28

FUZZY SYSTEMS .................................................................. 28NOISE REDUCTION BY FUZZY IMAGE FILTERING ...................................................................................28


3/32

JOURNAL AND CONFERENCE PAPERS .................................... 28A SEMI FRAGILE CONTENT BASED IMAGE WATERMARKING FOR AUTHENTICATION IN SPATIAL DOMAINUSING DISCRETE COSINE TRANSFORM ................................................................................................29PREDICTIVE JOB SCHEDULING IN A CONNECTION LIMITED SYSTEM USING PARALLEL GENETICALGORITHM ...........................................................................................................................................29PERSONALIZED WEB SEARCH WITH SELF-ORGANIZING MAP ...............................................................29ENHANCING PRIVACY AND AUTHORIZATION CONTROL SCALABILITY IN THE GRID THROUGHONTOLOGIES ...................................................................................................................... ...... .............29SPREAD SPECTRUM WATERMARKING SECURITY ...................................................................................30DOUBLE-COVERED BROADCAST (DCB): A SIMPLE RELIABLE BROADCAST ALGORITHM IN MANETS ....30COLLUSIVE PIRACY PREVENTION IN P2P CONTENT DELIVERY NETWORKS ...........................................30COMBINATORIAL APPROACH FOR PREVENTING SQL INJECTION ATTACKS ...........................................31A DISTRIBUTED AND SCALABLE ROUTING TABLE MANAGER FOR THE NEXT GENERATION OF IPROUTERS ....................................................................................................................................... ...... ..31A NOVEL SECURE COMMUNICATION PROTOCOL FOR AD HOC NETWORKS [SCP] ................................32IMAGE PROCESSING FOR EDGE DETECTION .........................................................................................32

NETWORKING

1

RESEQUENCING ANALYSISOF STOP-AND-WAITARQ FOR PARALLELMULTICHANNELCOMMUNICATIONS

DOTNET

In this paper, we consider a multichannel data communication syst

which the stop-and-wait automatic-repeat request protocol for pchannels with an in-sequence delivery guarantee (MSW-ARQ-inS) is userror control. We evaluate the resequencing delay and the resequebuffer occupancy, respectively. Under the assumption that all chahave the same transmission rate but possibly different time-invariantrates, we derive the probability generating function of the resequebuffer occupancy and the probability mass function of the resequedelay. Then, by assuming the GilbertElliott model for each channeextend our analysis to time-varying channels. Through examplecompute the probability mass functions of the resequencing occupancy and the resequencing delay for time-invariant channels. numerical and simulation results, we analyze trends in the resequencing buffer occupancy and the mean resequencing delfunctions of system parameters. We expect that the modeling techand analytical approach used in this paper can be applied tperformance evaluation of other ARQ protocols (e.g., the selective-r

ARQ) over multiple time-varying channels. Index TermsIn-seqdelivery, modeling and performance, multichannel data communicaresequencing buffer occupancy, resequencing delay, SW-ARQ.

2

RESOURCE ALLOCATIONIN OFDMAWIRELESSCOMMUNICATIONSSYSTEMSSUPPORTINGMULTIMEDIASERVICES

DOTNET

We design a resource allocation algorithm for down-link of orthofrequency division multiple access (OFDMA) systems supporting rea(RT) and best-effort (BE) services simultaneously over a time-vwireless channel. The proposed algorithm aims at maximizing sthroughput while satisfying quality of service (QoS) requirements of tand BE services. We take two kinds of QoS requirements into accountis the required average transmission rate for both RT and BE serviceother is the tolerable average absolute deviation of transmission(AADTR) just for the RT services, which is used to control the fluctuatransmission rates and to limit the RT packet delay to a moderate leveformulate the optimization problem representing the resource allounder consideration and solve it by using the dual optimization techand the projection stochastic subgradient method. Simulation resultsthat the proposed algorithm well meets the QoS requirements with the

throughput and outperforms the modified largest weighted delay firLWDF) algorithm that supports similar QoS requirements.


4/32

3

ANALYSIS OF SHORTESTPATH ROUTING FORLARGE MULTI-HOPWIRELESSNETWORKS

DOTNET

In this paper, we analyze the impact of straight line routing in homogeneous multi-hop wireless networks. We estimate the nodalwhich is defined as the number of packets served at a node, inducstraight line routing. For a given total offered load on the networkanalysis shows that the nodal load at each node is a function of the nVoronoi cell, the nodes location in the network, and the traffic paspecified by the source and destination randomness and straighrouting. In the asymptotic regime, we show that each nodes proba

that the node serves a packet arriving to the network approacheproducts of half the length of the Voronoi cell perimeter and thedensity function that a packet goes through the nodes location. The defunction depends on the traffic pattern generated by straight line roand determines where the hot spot is created in the network. Hcontrary to conventional wisdom, straight line routing can balance theover the network, depending on the traffic patterns.

4

SECURE AND POLICY-

COMPLIANTSOURCE ROUTING

DOT

NET

In todays Internet, inter-domain route control remains enevertheless, such control could improve the performance, reliabilityutility of the network for end users and ISPs alike. While researchersproposed a number of source routing techniques to combat this limitthere has thus far been no way for independent ASes to ensure thattraffic does not circumvent local traffic policies, nor to accurately detethe correct party to charge for forwarding the traffic. We present Plaan authenticated source routing system built around the concept of necapabilities, which allow for accountable, fine-grained path selecticryptographically attesting to policy compliance at each hop along a sroute. Capabilities can be composed to construct routes through muASes and can be delegated to third parties. Platypus caters to the neboth end users and ISPs: users gain the ability to pool their resourceselect routes other than the default, while ISPs maintain control over wwhen, and whose packets traverse their networks. We describe the dand implementation of an extensive Platypus policy framework that cused to address several issues in wide-area routing at both the edgthe core, and evaluate its performance and security. Our results showincremental deployment of Platypus can achieve immediate gains.

5

NODE ISOLATION MODELAND AGE-BASED

NEIGHBORSELECTION INUNSTRUCTUREDP2P NETWORKS

JAVA

Previous analytical studies of unstructured P2P resilience have assexponential user lifetimes and only con-sidered age-independent neireplacement. In this paper, we overcome these limitations by introducgeneral node-isolation model for heavy-tailed user lifetimes and arbneighbor-se-lection algorithms. Using this model, we analyze twobiased neighbor-selection strategies and show that they signifiimprove the residual lifetimes of chosen users, which dramatically re

the probability of user isolation and graph partitioning compareduniform selection of neighbors. In fact, the second strategy baserandom walks on age-proportional graphs demonstrates that, for lifewith infinite variance, the system monotonically increases its resilienits age and size grow. Specifically, we show that the probability of isoconverges to zero as these two metrics tend to infinity. We finish the with simulations in finite-size graphs that demonstrate the effect oresult in practice.

6

MOBILITY MANAGEMENTAPPROACHES FORMOBILE IP

NETWORKS:PERFORMANCECOMPARISON ANDUSERECOMMENDATIONS

JAVA

In wireless networks, efficient management of mobility is a crucial isssupport mobile users. The Mobile Internet Protocol (MIP) has been proto support global mobility in IP networks. Several mobility managstrategies have been proposed which aim reducing the signaling related to the Mobile Terminals (MTs) registration with the Home A(HAs) whenever their Care-of-Addresses (CoAs) change. They use difForeign Agents (FAs) and Gateway FAs (GFAs) hierarchies to concenthe registration processes. For high-mobility MTs, the Hierarchica

(HMIP) and Dynamic HMIP (DHMIP) strategies localize the registration and GFAs, yielding to high-mobility signaling. The Multicast HMIP strlimits the registration processes in the GFAs. For high-mobility Mprovides lowest mobility signaling delay compared to the HMIP and Dapproaches. However, it is resource consuming strategy unless for freMT mobility. Hence, we propose an analytic model to evaluate the signaling delay and the mean bandwidth per call according to the tyMT mobility. In our analysis, the MHMIP outperforms the DHMIP anstrategies in almost all the studied cases. The main contribution opaper is the analytic model that allows the mobility manageapproaches performance evaluation.


5/32

7

SINGLE-LINK FAILUREDETECTION IN ALL-

OPTICALNETWORKS USINGMONITORINGCYCLES AND PATHS

DOTNET

In this paper, we consider the problem of fault localization in all-onetworks. We introduce the concept of monitoring cycles (MCsmonitoring paths (MPs) for unique identification of single-link failuresand MPs are required to pass through one or more monitoring loca

They are constructed such that any single-link failure results in the faia unique combination of MCs and MPs that pass through the monilocation(s). For a network with only one monitoring location, we provthree-edge connectivity is a necessary and sufficient conditio

constructing MCs that uniquely identify any single-link failure inetwork. For this case, we formulate the problem of constructing MCs integer linear program (ILP). We also develop heuristic approachconstructing MCs in the presence of one or more monitoring locationan arbitrary network (not necessarily three-edge connected), we descfault localization technique that uses both MPs and MCs and that emmultiple monitoring locations. We also provide a linear-time algoritcompute the minimum number of required monitoring locations. Thextensive simulations, we demonstrate the effectiveness of the promonitoring technique.

8

MULTIPLE ROUTINGCONFIGURATIONSFOR FAST IPNETWORKRECOVERY

JAVA

As the Internet takes an increasingly central role in our communicainfrastructure, the slow convergence of routing protocols after a nefailure becomes a growing problem. To assure fast recovery from linnode failures in IP networks, we present a new recovery scheme Multiple Routing Configurations (MRC). Our proposed scheme guararecovery in all single failure scenarios, using a single mechanism to hboth link and node failures, and without knowing the root cause ofailure. MRC is strictly connectionless, and assumes only destination hop-by-hop forwarding. MRC is based on keeping additional rinformation in the routers, and allows packet forwarding to continue alternative output link immediately after the detection of a failure. It cimplemented with only minor changes to existing solutions. In this pappresent MRC, and analyze its performance with respect to scalabackup path lengths, and load distribution after a failure. We also showan estimate of the traffic demands in the network can be used to imthe distribution of the recovered traffic, and thus reduce the chanccongestion when MRC is used.

9 VIRUS SPREAD INNETWORKS

DOTNET

We study how the spread of computer viruses, worms, and othereplicating malware is affected by the logical topology of the networwhich they propagate. We consider a model in which each host canone of 3 possible states - susceptible, infected or removed (cured alonger susceptible to infection). We characterize how the size opopulation that eventually becomes infected depends on the ne

topology. Specially, we show that if the ratio of cure to infection ralarger than the spectral radius of the graph, and the initial inpopulation is small, then the final infected population is also smasense that can be made precise. Conversely, if this ratio is smaller thaspectral radius, then we show in some graph models of practical in(including power law random graphs) that the final infected populatlarge. These results yield insights into what the critical parameters determining virus spread in networks.

10

PRESTO:FEEDBACKDRIVEN

DATAMANAGEMENT INSENSORNETWORKS

DOTNET

This paper presents PRESTO, a novel two-tier sensor data managarchitecture comprising proxies and sensors that cooperate withanother for acquiring data and processing queries. PRESTO pconstruct time-series models of observed trends in the sensor dattransmit the parameters of the model to sensors. Sensors check sedata with model-predicted values and transmit only deviations fropredictions back to the proxy. Such a model-driven push approaenergyefficient, while ensuring that anomalous data trends are

missed. In addition to supporting queries on current data, PRESTOsupports queries on historical data using interpolation and local archsensors. PRESTO can adapt model and system parameters to datquery dynamics to further extract energy savings. We have implemPRESTO on a sensor testbed comprising Intel Stargates and Telos MOur experiments show that in a temperature monitoring applicPRESTO yields one to two orders of magnitude reduction in erequirements over on-demand, proactive or model-driven pull approaPRESTO also results in an order of magnitude reduction in query latea 1% duty-cycled five hop sensor network over a system that forwarqueries to remote sensor nodes.


6/32

11

EXPLICIT LOADBALANCINGTECHNIQUE FORNGEO SATELLITE IPNETWORKS WITHON-BOARDPROCESSINGCAPABILITIES

DOTNET

Non-geostationary (NGEO) satellite communication systems offer an of advantages over their terrestrial and geostationary counterparts.are seen as an integral part of next generation ubiquitous communicsystems. Given the non-uniform distribution of users in satellite footdue to several geographical and/or climatic constraints, some Inter-SaLinks (ISLs) are expected to be heavily loaded with data packets others remain underutilized. Such scenario obviously leads to congestthe heavily loaded links. It ultimately results in buffer overflows, h

queuing delays, and significant packet drops. To guarantee a distribution of traffic among satellites, this paper proposes an eexchange of information on congestion status among neighboring sateIndeed, a satellite notifies its congestion status to its neighboring sateWhen it is about to get congested, it requests its neighboring satellidecrease their data forwarding rates by sending them a self notification signaling message. In response, the neighboring satsearch for less congested paths that do not include the satellite in quand communicate a portion of data, primarily destined to the satellitthe retrieved paths. This operation avoids both congestion and packet at the satellite. It also ensures a better distribution of traffic over the satellite constellation. The proposed scheme is dubbed Explicit Balancing (ELB) scheme. While the multi-path routing concept of ELmany advantages, it may lead to persistent packet reordering. In caconnection- oriented protocols, this phenomenon results in unneceshrinkage of the data transmission rate. A solution to this issue i

incorporated in the design of ELB. The interactions of ELB with mechathat provide different QoS by differentiating traffic (e.g., DifferenServices) are also discussed. The good performance of ELB, in terbetter traffic distribution, higher throughput, and lower packet droverified via a set of simulations using the Network Simulator (NS).

12

DELAY ANALYSIS FORMAXIMALSCHEDULING WITHFLOW CONTROL INWIRELESSNETWORKS WITHBURSTY TRAFFIC

DOTNET

2009

We consider the delay properties of one-hop networks with ginterference constraints and multiple traffic streams with time-correarrivals. We first treat the case when arrivals are modulateindependent finite state Markov chains. We show that the well kmaximal scheduling algorithm achieves average delay that grows atlogarithmically in the largest number of interferers at any link. Furththe important special case when each Markov process has at mosstates (such as bursty ON/OFF sources), we prove that average deindependent of the number of nodes and links in the network, and heorder-optimal. We provide tight delay bounds in terms of the indiauto-correlation parameters of the traffic sources. These are perhapfirst order-optimal delay results for controlled queueing networksexplicitly account for such statistical information. Our analysis treats both with and without flow control.

13

MINING FILEDOWNLOADINGTIME INSTOCHASTIC PEERTO PEERNETWORKS

DOTNET

On-demand routing protocols use route caches to make routing deciDue to mobility, cached routes easily become stale. To address the staleness issue, prior work in DSR used heuristics with ad hoc parametpredict the lifetime of a link or a route. However, heuristics caccurately estimate timeouts because topology changes are unpredicIn this paper, we propose proactively disseminating the brokeninformation to the nodes that have that link in their caches. We defnew cache structure called a cache table and present a distributed cupdate algorithm. Each node maintains in its cache table the informnecessary for cache updates. When a link failure is detected, the algonotifies all reachable nodes that have cached the link in a distrmanner. The algorithm does not use any ad hoc parameters, thus mroute caches fully adaptive to topology changes. We show thaalgorithm outperforms DSR with path caches and with Link-Max Liadaptive timeout mechanism for link caches. We conclude that proacache updating is key to the adaptation of on-demand routing protocmobility.


7/32

1

4

ENHANCING SEARCHPERFORMANCE INUNSTRUCTURED

P2P NETWORKSBASED ON USERS'COMMON INTEREST

JAVA

Peer-to-peer (P2P) networks establish loosely coupled applicationoverlays on top of the Internet to facilitate efficient sharing of resou

They can be roughly classified as either structured or unstrunetworks. Without stringent constraints over the network topunstructured P2P networks can be constructed very efficiently antherefore considered suitable to the Internet environment. Howeverandom search strategies adopted by these networks usually pepoorly with a large network size. In this paper, we seek to enhanc

search performance in unstructured P2P networks through exploiting common interest patterns captured within a probability-theoretic framtermed the user interest model (UIM). A search protocol and a routingupdating protocol are further proposed in order to expedite the sprocess through self organizing the P2P network into a small worldtheoretical and experimental analyses are conducted and demonstrateeffectiveness and efficiency of our approach

15

BRA: A BIDIRECTIONALROUTINGABSTRACTION FORASYMMETRIC

MOBILE AD HOCNETWORKS

JAVA

Wireless links are often asymmetric due to heterogeneity intransmission power of devices, non-uniform environmental noise, andsignal propagation phenomenons. Unfortunately, routing protocomobile ad hoc networks typically work well only in bidirectional netw

This paper first presents a simulation study quantifying the impasymmetric links on network connectivity and routing performance. Itpresents a framework called BRA that provides a bidirectional abstractthe asymmetric network to routing protocols. BRA works by maintamulti-hop reverse routes for unidirectional links and provides threeabilities: improved connectivity by taking advantage of the unidireclinks, reverse route forwarding of control packets to enable off-therouting protocols, and detection packet loss on unidirectional Extensive simulations of AODV layered on BRA show that packet deincreases substantially (two-fold in some instances) in asymmetric netcompared to regular AODV, which only routes on bidirectional links.

16

RATE & DELAY GUARANTEESPROVIDED BY CLOSE PACKET

SWITCHES WITHLOAD BALANCING

JAVA

In this paper, we consider an overarching problem that encompassesperformance metrics. In particular, we study the network capacity prunder a given network lifetime requirement. Specifically, for a wsensor network where each node is provisioned with an initial energynodes are required to live up to a certain lifetime criterion, Sincobjective of maximizing the sum of rates of all the nodes in the netwolead to a severe bias in rate allocation among the nodes, we advocause of lexicographical max-min (LMM) rate allocation. To calculate therate allocation vector, we develop a polynomial-time algorithm by expthe parametric analysis (PA) technique from linear program (LP), whicall serial LP with Parametric Analysis (SLP-PA). We show that the S

can be also employed to address the LMM node lifetime problem more efficiently than a state-of-the-art algorithm proposed in the literMore important, we show that there exists an elegant duality relatiobetween the LMM rate allocation problem and the LMM node lifproblem. Therefore, it is sufficient to solve only one of the two probImportant insights can be obtained by inferring duality results for the problem.

17 GEOMETRIC APPROACH

TO IMPROVINGACTIVE PACKETLOSSMEASUREMENT

JAVA Measurement and estimation of packet loss characteristics are challedue to the relatively rare occurrence and typically short duration of ploss episodes. While active probe tools are commonly used to mepacket loss on end-to-end paths, there has been little analysis oaccuracy of these tools or their impact on the network. The objective study is to understand how to measure packet loss episodes accuratelend-to-end probes. We begin by testing the capability of standard Pomodulated end-to-end measurements of loss in a controlled laboenvironment using IP routers and commodity end hosts. Our tests show

loss characteristics reported from such Poisson-modulated probe toobe quite inaccurate over a range of traffic conditions. Motivated by observations, we introduce a new algorithm for packet loss measurethat is designed to overcome the deficiencies in standard Poisson-tools. Specifically, our method entails probe experiments that fogeometric distribution to 1) enable an explicit trade-off between accand impact on the network, and 2) enable more accurate measuremthan standard Poisson probing at the same rate. We evaluatcapabilities of our methodology experimentally by developingimplementing a prototype tool, called BADABING. The experidemonstrate the trade-offs between impact on the networkmeasurement accuracy. We show that BADABING reports


8/32

characteristics far more accurately than traditional loss measurement

18

PERFORMANCE OF ASPECULATIVETRANSMISSIONSCHEME FORSCHEDULINGLATENCY

REDUCTION

JAVA

This work was motivated by the need to achieve low latency in an centrally-scheduled cell switch for high-performance comapplications; specifically, the aim is to reduce the latency incurred betissuance of a request and arrival of the corresponding grant. We introdspeculative transmission scheme to significantly reduce the avlatency by allowing cells to proceed without waiting for a grant. It opin conjunction with any centralized matching algorithm to achieve amaximum utilization. An analytical model is presented to investigatefficiency of the speculative transmission scheme employed in ablocking N*NR input-queued crossbar switch with receivers R per o

The results demonstrate that the can be almost entirely eliminated forup to 50%. Our simulations confirm the analytical results.

19

RATE ALLOCATION &NETWORK LIFETIMEPROBLEM FORWIRELESS SENSORNETWORKS

DOTNET

In this paper, we consider an overarching problem that encompassesperformance metrics. In particular, we study the network capacity prunder a given network lifetime requirement. Specifically, for a wsensor network where each node is provisioned with an initial energynodes are required to live up to a certain lifetime criterion, Sincobjective of maximizing the sum of rates of all the nodes in the netwolead to a severe bias in rate allocation among the nodes, we advocause of lexicographical max-min (LMM) rate allocation. To calculate therate allocation vector, we develop a polynomial-time algorithm by expthe parametric analysis (PA) technique from linear program (LP), whicall serial LP with Parametric Analysis (SLP-PA). We show that the Scan be also employed to address the LMM node lifetime problem more efficiently than a state-of-the-art algorithm proposed in the literMore important, we show that there exists an elegant duality relatio

between the LMM rate allocation problem and the LMM node lifproblem. Therefore, it is sufficient to solve only one of the two probImportant insights can be obtained by inferring duality results for the problem.

20 STATISTICAL TECHNIQUES

FOR DETECTINGTRAFFICANOMALIESTHROUGH PACKETHEADER DATA

DOTNET

THE frequent attacks on network infrastructure, using various fordenial of service (DoS) attacks and worms, have led to an increasedfor developing techniques for analyzing and monitoring network traefficient analysis tools were available, it could become possible to dthe attacks, anomalies and take action to suppress them before theyhad much time to propagate across the network. In this paper, we studpossibilities of traffic-analysis based mechanisms for attack and anodetection. The motivation for this work came from a need to reduclikelihood that an attacker may hijack the campus machines to staattack on a third party. A campus may want to prevent or limit misusemachines in staging attacks, and possibly limit the liability fromattacks. In particular, we study the utility of observing packet heade

of outgoing traffic, such as destination addresses, port numbers annumber of flows, in order to detect attacks/anomalies originating frocampus at the edge of a campus. Detecting anomalies/attacks close tsource allows us to limit the potential damage close to the attamachines. Traffic monitoring close to the source may enable the neoperator quicker identification of potential anomalies and allow control of administrative domains resources. Attack propagation couslowed through early detection. Our approach passively monitors netraffic at regular intervals and analyzes it to find any abnormalities aggregated traffic. By observing the traffic and correlating it to prestates of traffic, it may be possible to see whether the current trabehaving in a similar (i.e., correlated) manner. The network traffic


9/32

look different because of flash crowds, changing access patinfrastructure problems such as router failures, and DoS attacks. In theof bandwidth attacks, the usage of network may be increasedabnormalities may show up in traffic volume. Flash crowds couobserved through sudden increase in traffic volume to a single destin

Sudden increase of traffic on a certain port could signify the onset anomaly such as worm propagation. Our approach relies on anapacket header data in order to provide indications of Possible abnormin the traffic.

21

EFFICIENT ROUTING ININTERMITTENTLYCONNECTEDMOBILENETWORKS: THEMULTIPLE COPYCASE

DOTNET

Intermittently connected mobile networks are wireless networks most of the time there does not exist a complete path from the sourthe destination. There are many real networks that follow this modexample, wildlife tracking sensor networks, military networks, vehicuhoc networks, etc. In this context, conventional routing schemebecause they try to establish complete end-to-end paths, before any dsent. To deal with such networks researchers have suggested tflooding-based routing schemes. While flooding-based schemes have probability of delivery, they waste a lot of energy and suffer from scontention which can significantly degrade their performance. Furtherproposed efforts to reduce the overhead of flooding-based schemesoften been plagued by large delays. With this in mind, we introduce family of routing schemes that spray a few message copies intnetwork, and then route each copy independently towards the destinWe show that, if carefully designed, spray routing

22

TWO TECHNIQUES FORFASTCOMPUTATION OFCONSTRAINEDSHORTEST PATHS

JAVA

Computing constrained shortest paths is fundamental to some impnetwork functions such as QoS routing, MPLS path selection, ATM crouting, and traffic engineering. The problem is to find the cheapestthat satisfies certain constraints. In particular, finding the cheapest dconstrained path is critical for real-time data flows such as voice/videoBecause it is NP-complete, much research has been designing healgorithms that solve the -approximation of the problem with an adjuaccuracy. A common approach is to discretize (i.e., scale and roundlink delay or link cost, which transforms the original problem to a sione solvable in polynomial time. The efficiency of the algorithms direlates to the magnitude of the errors introduced during discretizat

this paper, we propose two techniques that reduce the discretization ewhich allow faster algorithms to be designed. Reducing the overhecomputing constrained shortest paths is practically important fosuccessful design of a high-throughput QoS router, which is limited aprocessing power and memory space. Our simulations show that thealgorithms reduce the execution time by an order of magnitude on plaw topologies with 1000 nodes.

23

PROBABILISTIC PACKETMARKING FORLARGE-SCALE IPTRACE BACK

DOTNET

An approach to IP traces back based on the probabilistic packet maparadigm. Our approach, which we call randomize-and-link, uses checksum cords to link message fragments in a way that is scalable, for the checksums serve both as associative addresses andintegrity verifiers. The main advantage of these checksum cords is thaspread the addresses of possible router messages across a spectrum ttoo large for the attacker to easily create messages that collidelegitimate messages.


10/32

24

DUAL-LINK FAILURERESILIENCYTHROUGH BACKUPLINK MUTUALEXCLUSION

JAVA

Networks employ link protection to achieve fast recovery from link faWhile the first link failure can be protected using link protection, theseveral alternatives for protecting against the second failure. This formally classifies the approaches to dual-link failure resiliency. One strategies to recover from dual-link failures is to employ link protectithe two failed links independently, which requires that two links mause each other in their backup paths if they may fail simultaneously. Srequirement is referred to as backup link mutual exclusion (

constraint and the problem of identifying a backup path for every linsatisfies the above requirement is referred to as the BLME problempaper develops the necessary theory to establish the sufficient condfor existence of a solution to the BLME problem. Solution methodologthe BLME problem is developed using two approaches by: 1) formuthe backup path selection as an integer linear program; 2) developpolynomial time heuristic based on minimum cost path routing. Thformulation and heuristic are applied to six networks and their performis compared with approaches that assume precise knowledge of dufailure. It is observed that a solution exists for all of the six netconsidered. The heuristic approach is shown to obtain feasible solthat are resilient to most dual-link failures, although the backuplengths may be significantly higher than optimal. In addition, the illustrates the significance of the knowledge of failure location by illustthat network with higher connectivity may require lesser capacity thawith a lower connectivity to recover from dual-link failures.

25

A DISTRIBUTEDDATABASEARCHITECTURE FORGLOBAL ROAMINGIN NEXT-GENERATIONMOBILE NETWORKS

JAVA

The next-generation mobile network will support terminal mobility, pemobility, and service provider portability, making global roaming seamA location-independent personal telecommunication number (PTN) scis conducive to implementing such a global mobile system. Howevenon-geographic PTNs coupled with the anticipated large number of musers in future mobile networks may introduce very large centrdatabases. This necessitates research into the design and performahigh-throughput database technologies used in mobile systems to ethat future systems will be able to carry efficiently the anticipated l

This paper proposes a scalable, robust, efficient location daarchitecture based on the location-independent PTNs. The proposedtree database architecture consists of a number of database subsyseach of which is a three-level tree structure and is connected to the oonly through its root. By exploiting the localized nature of callinmobility patterns, the proposed architecture effectively reducedatabase loads as well as the signaling traffic incurred by the locregistration and call delivery procedures. In addition, two memory-re

database indices, memory-resident direct file and T-tree, are proposthe location databases to further improve their throughput. Analysis and numerical results are presented to evaluate the efficiency oproposed database architecture. Results have revealed that the prodatabase architecture for location management can effectively suppoanticipated high user density in the future mobile networks.

26

NETWORK BORDERPATROL:PREVENTINGCONGESTIONCOLLAPSE AND

PROMOTINGFAIRNESS IN THEINTERNET

JAVA

The Internet's excellent scalability and robustness result in part froend-to-end nature of Internet congestion control. End-to-end congcontrol algorithms alone, however, are unable to prevent the congcollapse and unfairness created by applications that are unresponsnetwork congestion. To address these maladies, we proposeinvestigate a novel congestion-avoidance mechanism called network bpatrol (NBP). NBP entails the exchange of feedback between routers borders of a network in order to detect and restrict unresponsive tflows before they enter the network, thereby preventing congestion the network. Moreover, NBP is complemented with the proposed enh

core-stateless fair queueing (ECSFQ) mechanism, which providebandwidth allocations to competing flows. Both NBP and ECSFQcompliant with the Internet philosophy of pushing complexity towaredges of the network whenever possible. Simulation results show thaeffectively eliminates congestion collapse and that, when combinedECSFQ, approximately max-min fair bandwidth allocations can be achfor competing flows.


11/32

MOBILE COMPUTING

1A TABU SEARCH

ALGORITHM FORCLUSTER BUILDINGIN WIRELESSSENSOR NETWORKS

DOTNET

The main challenge in wireless sensor network deployment pertaoptimizing energy consumption when collecting data from sensor n

This paper proposes a new centralized clustering method for acollection mechanism in wireless sensor networks, which is base

network energy maps and Quality-of-Service (QoS) requirementsclustering problem is modeled as a hypergraph partitioning anresolution is based on a tabu search heuristic. Our approach defines musing largest size cliques in a feasibility cluster graph. Compared to methods (CPLEX-based method, distributed method, simulated annebased method), the results show that our tabu search-based appreturns high-quality solutions in terms of cluster cost and execution tima result, this approach is suitable for handling network extensibilitsatisfactory manner.

2

ROUTE STABILITY INMANETS UNDERTHE RANDOM

DIRECTIONMOBILITY MODEL

DOTNET

A fundamental issue arising in mobile ad hoc networks (MANETs) selection of the optimal path between any two nodes. A method thabeen advocated to improve routing efficiency is to select the most path so as to reduce the latency and the overhead due to reconstruction. In this work, we study both the availability and the duprobability of a routing path that is subject to link failures caused bymobility. In particular, we focus on the case where the network nodes

according to the Random Direction model, and we derive both exacapproximate (but simple) expressions of these probabilities. Througresults, we study the problem of selecting an optimal route in terms ofavailability. Finally, we propose an approach to improve the efficienreactive routing protocols.

3

GREEDY ROUTING WITH

ANTI-VOIDTRAVERSAL FORWIRELESS SENSORNETWORKS

DOTNET

The unreachability problem (i.e., the so-called void problem) that exthe greedy routing algorithms has been studied for the wireless snetworks. Some of the current research work cannot fully resolve theproblem, while there exist other schemes that can guarantee the delivpackets with the excessive consumption of control overheads. In this a greedy antivoid routing (GAR) protocol is proposed to solve theproblem with increased routing efficiency by exploiting the boufinding technique for the unit disk graph (UDG). The proposed rollinUDG boundary traversal (RUT) is employed to completely guarantedelivery of packets from the source to the destination node under thenetwork. The boundary map (BM) and the indirect map searchingscheme are proposed as efficient algorithms for the realization of th

technique. Moreover, the hop count reduction (HCR) scheme is utilizeshort-cutting technique to reduce the routing hops by listening tneighbors traffic, while the intersection navigation (IN) mechaniproposed to obtain the best rolling direction for boundary traversal wiadoption of shortest path criterion. In order to maintain the nerequirement of the proposed RUT scheme under the non-UDG networkpartial UDG construction (PUC) mechanism is proposed to transfornon-UDG into UDG setting for a portion of nodes that facilitate boutraversal. These three schemes are incorporated within the GAR protofurther enhance the routing performance with reduced communioverhead. The proofs of correctness for the GAR scheme are also givthis paper. Comparing with the existing localized routing algorithmsimulation results show that the proposed GAR-based protocols can prbetter routing efficiency.


12/32

4

CELL BREATHING

TECHNIQUES FORLOAD BALANCINGIN WIRELESS LANS

DOTNET

Maximizing network throughput while providing fairness is one of thchallenges in wireless LANs (WLANs). This goal is typically achieved the load of access points (APs) is balanced. Recent studies on operaWLANs, however, have shown that AP load is often substantially unevalleviate such imbalance of load, several load balancing schemesbeen proposed. These schemes commonly require proprietary softwhardware at the user side for controlling the user-AP association. Ipaper we present a new load balancing technique by controlling the s

WLAN cells (i.e., APs coverage range), which is conceptually similar tbreathing in cellular networks. The proposed scheme does not requirmodification to the users neither the IEEE 802.11 standard. It only reqthe ability of dynamically changing the transmission power of thbeacon messages. We develop a set of polynomial time algorithms thathe optimal beacon power settings which minimize the load of thecongested AP. We also consider the problem of network-wide min-maxbalancing. Simulation results show that the performance of the promethod is comparable with or superior to the best existing associbased methods.

5CAR: CONTEXT-AWARE

ADAPTIVE ROUTINGFOR DELAY-TOLERANT MOBILENETWORKS

JAVA

Most of the existing research work in mobile ad hoc networking is basthe assumption that a path exists between the sender and the receivethe other hand, applications of decentralised mobile systems are characterised by network partitions. As a consequence delay tonetworking research has received considerable attention in the recent as a means to obviate to the gap between ad hoc network researcreal applications. In this paper we present the design, implementatioevaluation of the context-aware adaptive routing (CAR) protocol for tolerant unicast communication in intermittently connected mobile anetworks. The protocol is based on the idea of exploiting nodes as caof messages among network partitions to achieve delivery. The chothe best carrier is made using Kalman filter based prediction technand utility theory. We discuss the implementation of CAR ovopportunistic networking framework, outlining possible applications general principles at the basis of the proposed approach. The largeperformance of the CAR protocol are evaluated using simulations basa social network founded mobility model, a purely random one andtraces from Dartmouth College.

6

LOCAL CONSTRUCTION OFNEAR-OPTIMALPOWER SPANNERSFOR WIRELESS AD-HOC NETWORKS

DOTNET

We present a local distributed algorithm that, given a wireless anetwork modeled as a unit disk graph U in the plane, constructs a ppower spanner of U whose degree is bounded by k and whose stretch is bounded by 1 + (2sin pi/k)p, where k ges 10 is an integer parametep isin [2, 5] is the power exponent constant. For the same degree bothe stretch factor of our algorithm significantly improves the previoubounds by Song et al. We show that this bound is near-optimal by prthat the slightly smaller stretch factor of 1 + (2sin pi/k+1)p is unattafor the same degree bound k. In contrast to previous algorithms foproblem, the presented algorithm is local. As a consequence, the algois highly scalable and robust. Finally, while the algorithm is efficieneasy to implement in practice, it relies on deep insights on the geomeunit disk graphs and novel techniques that are of independent interest

7

BIASED RANDOM WALKS

IN UNIFORMWIRELESSNETWORKS

DOTNET

A recurrent problem when designing distributed applications is to seara node with known property. File searching in peer-to-peer applications, resource discovery in service-oriented architectures (Sand path discovery in routing can all be cast as a search problem. Rawalk-based search algorithms are often suggested for tackling the sproblem, especially in very dynamic systems-like mobile wireless netw

The cost and the effectiveness of a random walk-based search algoare measured by the excepted number of transmissions required bhitting the target. Hence, to have a low hitting time is a critical goapaper studies the effect of biasing random walk toward the target ohitting time. For a walk running over a network with uniform distribution, a simple upper bound that connects the hitting time to thlevel is obtained. The key result is that even a modest bias level is areduce the hitting time significantly. This paper also proposes a sprotocol for mobile wireless networks, whose results are interpreted light of the theoretical study. The proposed solution is for unstrucwireless mobile networks.


13/32

8

INFORMATION CONTENT-BASED SENSORSELECTION AND

TRANSMISSIONPOWERADJUSTMENT FORCOLLABORATIVETARGET TRACKING

DOT

NET

For target tracking applications, wireless sensor nodes provide accinformation since they can be deployed and operated neaphenomenon. These sensing devices have the opportunity of collaboamong themselves to improve the target localization and traaccuracies. An energy-efficient collaborative target tracking paradideveloped for wireless sensor networks (WSNs). A mutual-informbased sensor selection (MISS) algorithm is adopted for participation fusion process. MISS allows the sensor nodes with the highest m

information about the target state to transmit data so that the econsumption is reduced while the desired target position estimaccuracy is met. In addition, a novel approach to energy savings in Wdevised in the information-controlled transmission power (ICTP) adjustwhere nodes with more information use higher transmission powersthose that are less informative to share their target state informatiothe neighboring nodes. Simulations demonstrate the performance offered by MISS and ICTP in terms of power consumption and localization accuracy.

9

ENERGY MAPS FORMOBILE WIRELESSNETWORKS:COHERE

NCE TIME VERSUSSPREADING PERIOD

DOTNET

We show that even though mobile networks are highly unpredictableviewed at the individual node scale, the end-toend quality-of-service metrics can be stationary when the mobile network is viewed iaggregate. We define the coherence time as the maximum duratiowhich the end-to-end QoS metric remains roughly constant, anspreading period as the minimum duration required to spreadinformation to all the nodes. We show that if the coherence time is gthan the spreading period, the end-to-end QoS metric can be trackefocus on the energy consumption as the end-to-end QoS metricdescribe a novel method by which an energy map can be constructerefined in the joint memory of the mobile nodes. Finally, we showenergy maps can be utilized by an application that aims to minimnodes total energy consumption over its near-future trajectory.

10

RANDOMCAST: ANENERGY EFFICIENTCOMMUNICATIONSCHEME FOR

MOBILE AD HOCNETWORKS

DOTNET

In mobile ad hoc networks (MANETs), every node overhears everytransmission occurring in its vicinity and thus, consumes eunnecessarily. In IEEE 802.11 Power Saving Mechanism (PSM), a pmust be advertised before it is actually transmitted. When a node rean advertised packet that is not destined to itself, it switches to a lowsleep state during the data transmission period, and thus, aoverhearing and conserves energy. However, since some MANET roprotocols such as Dynamic Source Routing (DSR) collect route informvia overhearing, they would suffer if they are used in combination802.11 PSM. Allowing no overhearing may critically deteriorateperformance of the underlying routing protocol, while uncond

overhearing may offset the advantage of using PSM. This paper proponew communication mechanism, called RandomCast, via which a scan specify the desired level of overhearing, making a prudent babetween energy and routing performance. In addition, it reduces redurebroadcasts for a broadcast packet and thus saves more energy. Extesimulation using ns-2 shows that RandomCast is highly energy-efcompared to conventional 802.11 as well as 802.11 PSM-based schemterms of total energy consumption, energy goodput and energy balanc

11

EFFICIENT RESOURCE

ALLOCATION FORWIRELESSMULTICAST

JAVA

In this paper, we propose a bandwidth-efficient multicast mechanisheterogeneous wireless networks. We reduce the bandwidth cost Internet Protocol (IP) multicast tree by adaptively selecting the cell anwireless technology for each mobile host to join the multicast groupmechanism enables more mobile hosts to cluster together and leads tuse of fewer cells to save the scarce wireless bandwidth. Besides, the in the multicast tree connecting to the selected cells share more comlinks to save the wireline bandwidth. Our mechanism supports the dyn

group membership and offers mobility of group members. Moreovemechanism requires no modification to the current IP multicast rprotocols. We formulate the selection of the cell and the wtechnology for each mobile host in the heterogeneous wireless netwoan optimization problem. We use Integer Linear Programming to modproblem and show that the problem is NP-hard. To solve the problempropose a distributed algorithm based on Lagrangean relaxation network protocol based on the algorithm. The simulation results showour mechanism can effectively save the wireless and wireline bandwidcompared to the traditional IP multicast.


14/32

12

INTRUSION DETECTION INHOMOGENEOUS &HETEROGENEOUSWIRELESS SENSORNETWORKS

JAVA

Intrusion detection in Wireless Sensor Network (WSN) is of practical inin many applications such as detecting an intruder in a battlefieldintrusion detection is defined as a mechanism for a WSN to detecexistence of inappropriate, incorrect, or anomalous moving attackethis paper, we consider this issue according to heterogeneous WSN mFurthermore, we consider two sensing detection models: single-sedetection and multiple-sensing detection... Our simulation results sho

advantage of multiple sensor heterogeneous WSNs.

13

LOCATION BASED SPATIALQUERYPROCESSING INWIRELESSBROADCASTENVIRONMENTS

JAVA

Location-based spatial queries (LBSQ s) refer to spatial queries wanswers rely on the location of the inquirer. Efficient processing of LBSof critical importance with the ever-increasing deployment and umobile technologies. We show that LBSQ s has certain ucharacteristics that the traditional spatial query processing in centrdatabases does not address. For example, a significant challenpresented by wireless broadcasting environments, which have excscalability but often exhibit high-latency database access. In this papepresent a novel query processing technique that, though maintainingscalability and accuracy, manages to reduce the latency consideraanswering LBSQ s. Our approach is based on peer-to-peer sharing, enables us to process queries without delay at a mobile host by using results cached in its neighboring mobile peers. We demonstratfeasibility of our approach through a probabilistic analysis, and we illuthe appeal of our technique through extensive simulation results.

14

BANDWIDTH ESTIMATIONFOR IEEE 802.11BASED ADHOCNETWORK

JAVA

Since 2005, IEEE 802.11-based networks have been able to provcertain level of quality of service (QoS) by the means of sdifferentiation, due to the IEEE 802.11e amendment. Howevemechanism or method has been standardized to accurately evaluatamount of resources remaining on a given channel. Such an evaluwould, however, be a good asset for bandwidth-constrained applicatiomultihop ad hoc networks, such evaluation becomes even more difConsequently, despite the various contributions around this research the estimation of the available bandwidth still represents one of theissues in this field. In this paper, we propose an improved mechaniestimate the available bandwidth in IEEE 802.11-based ad hoc netw

Through simulations, we compare the accuracy of the estimatipropose to the estimation performed by other state-of-the-artprotocols, BRuIT, AAC, and QoS-AODV.

15

BENEFIT-BASED DATACACHING IN AD HOCNETWORKS

JAVA

Data caching can significantly improve the efficiency of information ain a wireless ad hoc network by reducing the access latency and band

usage. However, designing efficient distributed caching algorithms istrivial when network nodes have limited memory. In this articlconsider the cache placement problem of minimizing total data accesin ad hoc networks with multiple data items and nodes with limited mecapacity. The above optimization problem is known to be NP-hard. Debenefit as the reduction in total access cost, we present a polynomiacentralized approximation algorithm that provably delivers a solution wbenefit is at least one-fourth (one-half for uniform-size data items) optimal benefit. The approximation algorithm is amenable to locdistributed implementation, which is shown via simulations to performto the approximation algorithm. Our distributed algorithm naturally exto networks with mobile nodes. We simulate our distributed algorithm a network simulator (ns2), and demonstrate that it significantly outperanother existing caching technique (by Yin and Cao [30]) in all impoperformance metrics. The performance differential is particularly lamore challenging scenarios, such as higher access frequency and smmemory.

16 LOCALIZED SENSOR AREA

COVERAGE WITHLOWCOMMUNICATIONOVERHEAD

DOTNET

We propose several localized sensor area coverage protocolheterogeneous sensors, each with arbitrary sensing and transmissionEach sensor has a time out period and listens to messages serespective nodes before the time out expires. Sensor nodes whose searea is not fully covered (or fully covered but with a disconnected active sensors) when the deadline expires decide to remain active foconsidered round and transmit an activity message announcing it. approach, sensor decides to sleep only if neighbor sensor is active covered. Covered nodes decide to sleep, with or without transmitwithdrawal message to inform neighbors about the status. After hefrom more neighbors, inactive sensors may observe that they becovered and may decide to alter their original decision and trans


15/32

retreat message.

17

EFFICIENT RESOURCEALLOCATION FORWIRELESSMULTICAST

DOTNET

In this paper, we propose a bandwidth-efficient multicast mechanisheterogeneous wireless networks. We reduce the bandwidth cost omulticast tree by adaptively selecting the cell and the wireless technfor each mobile host to join the multicast group. Our mechanism enmore mobile hosts to cluster together and lead to the use of fewer csave the scarce wireless bandwidth. Besides, the paths in the multicasconnecting to the selected cells share more common links to savwireline bandwidth. Our mechanism supports the dynamic membership and offers mobility of group members. Moreovermechanism requires no modification on the current IP multicast roprotocols. We formulate the selection of the cell and the wtechnology for each mobile host in the heterogeneous wireless netwoan optimization problem. We use Integer Linear Programming to modproblem and show that the problem is NP-hard. To solve the problem

propose an distributed algorithm based on Lagrangean relaxation network protocol based on the algorithm. The simulation results showour mechanism can effectively save the wireless and wireline bandwidcompared to the traditional IP multicast.

18

AN ACKNOWLEDGMENT-BASED APPROACHFOR THEDETECTION OFROUTINGMISBEHAVIOR INMANETS

JAVA

We study routing misbehavior in MANETs (Mobile Ad Hoc Networks) paper. In general, routing protocols for MANETs are designed based oassumption that all participating nodes are fully cooperative. Howeveto the open structure and scarcely available battery-based energy,misbehaviors may exist. One such routing misbehavior is that somenodes will participate in the route discovery and maintenance processerefuse to forward data packets. In this paper, we propose the 2ACK scthat serves as an add-on technique for routing schemes to detect romisbehavior and to mitigate their adverse effect. The main idea of the scheme is to send two-hop acknowledgment packets in the opdirection of the routing path. In order to reduce additional routing overonly a fraction of the received data packets are acknowledged in the

scheme. Analytical and simulation results are presented to evaluatperformance of the proposed scheme.

19

ROUTE RESERVATION INAD HOC WIRELESSNETWORKS

JAVA

This paper investigates whether and when route reservation-basedcommunication can yield better delay performance than non-reservbased (NRB) communication in ad hoc wireless networks. In additposing this fundamental question, the requirements (in terms of discovery, medium access control (MAC) protocol, and pipelining, etmaking RB switching superior to NRB switching are also identified. A analytical framework is developed and the network performance undeRB and NRB schemes is quantified. It is shown that if the aforementrequirements are met, then RB schemes can indeed yield better performance than NRB schemes. This advantage, however, comes aexpense of lower throughput and goodput compared to NRB schemes.


16/32

20

DISTRIBUTED CACHEUPDATING FOR THE

DYNAMIC SOURCEROUTINGPROTOCOL

JAVA

On-demand routing protocols use route caches to make routing deciDue to mobility, cached routes easily become stale. To address the staleness issue, prior work in DSR used heuristics with ad hoc parametpredict the lifetime of a link or a route. However, heuristics caccurately estimate timeouts because topology changes are unpredicIn this paper, we propose proactively disseminating the brokeninformation to the nodes that have that link in their caches. We defnew cache structure called a cache table and present a distributed c

update algorithm. Each node maintains in its cache table the informnecessary for cache updates. When a link failure is detected, the algonotifies all reachable nodes that have cached the link in a distrmanner. The algorithm does not use any ad hoc parameters, thus mroute caches fully adaptive to topology changes. We show thaalgorithm outperforms DSR with path caches and with Link-Max Liadaptive timeout mechanism for link caches. We conclude that proacache updating is key to the adaptation of on-demand routing protocmobility.

PARALLEL AND DISTRIBUTED SYSTEMS

1DYNAMIC SEARCHALGORITHM IN

UNSTRUCTUREDPEER-TO-PEERNETWORKS

JAVA

Designing efficient search algorithms is a key challenge in unstrucpeer-to-peer networks. Flooding and random walk (RW) are two tsearch algorithms. Flooding searches aggressively and covers the nodes. However, it generates a large amount of query messages and,does not scale. On the contrary, RW searches conservatively. Itgenerates a fixed amount of query messages at each hop but wouldlonger search time. We propose the dynamic search (DS) algorithm, wh

a generalization of flooding and RW. DS takes advantage of various counder which each previous search algorithm performs well. It reseflooding for short-term search and RW for long-term search. Moreovecould be further combined with knowledge-based search mechanisimprove the search performance. We analyze the performance of DS on some performance metrics including the success rate, search time, hits, query messages, query efficiency, and search efficiency. Numresults show that DS provides a good tradeoff between search performand cost. On average, DS performs about 25 times better than floodin58 times better than RW in power-law graphs, and about 186 times than flooding and 120 times better than RW in bimodal topologies.

2FLEXIBLE DETERMINISTIC

PACKET MARKING:AN IP TRACEBACK

SYSTEM TO FINDTHE REAL SOURCEOF ATTACKS

JAVA Internet Protocol (IP) traceback is the enabling technology to cInternet crime. In this paper, we present a novel and practical IP tracsystem called Flexible Deterministic Packet Marking (FDPM) which proa defense system with the ability to find out the real sources of attapackets that traverse through the network. While a number of

traceback schemes exist, FDPM provides innovative features to tracsource of IP packets and can obtain better tracing capability than othparticular, FDPM adopts a flexible mark length strategy to macompatible to different network environments; it also adaptively changmarking rate according to the load of the participating router by a fleflow-based marking scheme. Evaluations on both simulation and real simplementation demonstrate that FDPM requires a moderately number of packets to complete the traceback process; add little addiload to routers and can trace a large number of sources in one traceprocess with low false positive rates. The built-in overload prevmechanism makes this system capable of achieving a satisfactory tracresult even when the router is heavily loaded. The motivation o


17/32

traceback system is from DDoS defense. It has been used to not onlyDDoS attacking packets but also enhance filtering attacking traffic. It

wide array of applications for other security systems.

3

DISTRIBUTEDALGORITHMS FORCONSTRUCTING

APPROXIMATEMINIMUMSPANNING TREESIN WIRELESSSENSORNETWORKS

JAVA

While there are distributed algorithms for the minimum spanning tree problem, these algorithms require relatively large number of messagetime, and are fairly involved, making them impractical for resoconstrained networks such as wireless sensor networks. In such netwosensor has very limited power, and any algorithm needs to be simple,and energy efficient. Motivated by these considerations, we desiganalyze a class of simple and local distributed algorithms called NeNeighbor Tree (NNT) algorithms for energy-efficient construction approximate MST in wireless networks. Assuming that the nodeuniformly distributed, we show provable bounds on both the quality spanning tree produced and the energy needed to construct them. Wethat while NNT produces a close approximation to the MST, it consasymptotically less energy than the classical message-optimal distriMST algorithm due to Gallagery, Humblet, and Spira. Further, the NNTbe maintained dynamically with polylogarithmic rearrangements underinsertions/deletions. We also perform extensive simulations, which that the bounds are much better in practice. Our results, to the best knowledge, demonstrate the first tradeoff between the qualiapproximation and the energy required for building spanning trewireless networks, and motivate similar considerations for other impproblems.

4

A FAITHFUL DISTRIBUTEDMECHANISM FORSHARING THE COSTOF MULTICASTTRANSMISSIONS

J2EE

The problem of sharing the cost of multicast transmissions was studthe past, and two mechanisms, Marginal Cost (MC) and Shapley Valuewere proposed to solve it. Although both of them are strategy mechanisms, the distributed protocols implementing them are susceto manipulation by autonomous nodes. We propose a distributed Sh

Value mechanism in which the participating nodes do not have incentivdeviate from the mechanism specifications. We show that the promechanism is a faithful implementation of the Shapley Value mechaWe experimentally investigate the performance of the existing anproposed cost-sharing mechanisms by implementing and deploying thePlanetLab. We compare the execution time of MC and SH mechanismthe Tamper-Proof and Autonomous Node models. We also studconvergence and scalability of the mechanisms by varying the numnodes and the number of users per node. We show that the MC mechagenerate a smaller revenue compared to the SH mechanisms, andthey are not attractive to the content provider. We also show that increthe number of users per node is beneficial for the systems implementinSH mechanisms from both computational and economic perspectives.

5DYNAMIC ROUTING WITH

SECURITYCONSIDERATIONS

JAVA

Security has become one of the major issues for data communicationwired and wireless networks. Different from the past work on the desicryptography algorithms and system infrastructures, we will prop

dynamic routing algorithm that could randomize delivery paths fortransmission. The algorithm is easy to implement and compatiblepopular routing protocols, such as the Routing Information Protocol in networks and Destination-Sequenced Distance Vector protocol in winetworks, without introducing extra control messages. An analytic stuthe proposed algorithm is presented, and a series of simulation experimare conducted to verify the analytic results and to show the capability proposed algorithm


18/32

6

COMPACTION OFSCHEDULES AND A

TWO-STAGEAPPROACH FORDUPLICATION-BASED DAGSCHEDULING

DOTNET

Many DAG scheduling algorithms generate schedules that rprohibitively large number of processors. To address this problempropose a generic algorithm, SC, to minimize the processor requiremany given valid schedule. SC preserves the schedule length of the orschedule and reduces processor count by merging processor scheduleremoving redundant duplicate tasks. To the best of our knowledge, tthe first algorithm to address this highly unexplored aspect ofscheduling. On the average, SC reduced the processor requirement

82% and 72% for schedules generated by PLW, TCSD and CPFD algorirespectively. SC algorithm has a low complexity (O(|N |3) ) comparmost duplication based algorithms. Moreover, it decouples proeconomization from schedule length minimization problem. Toadvantage of these features of SC, we also propose a scheduling algoSDS, having the same time complexity as SC. Our experiments demonthat, schedules generated by SDS are only 3% longer than CPFD (O(|None of the best algorithms in that respect. SDS and SC together form astage scheduling algorithm that produces schedules with high qualitlow processor requirement, and has lower complexity than the compalgorithms that produce similar high quality results.

7 DETECTING MALICIOUSPACKET LOSSES

JAVA

In this paper, we consider the problem of detecting whether a comprorouter is maliciously manipulating its stream of packets. In particular, wconcerned with a simple yet effective attack in which a router selecdrops packets destined for some victim. Unfortunately, it is challenging to attribute a missing packet to a malicious action becnormal network congestion can produce the same effect. Modern netroutinely drop packets when the load temporarily exceeds their bufcapacities. Previous detection protocols have tried to address this prwith a user-defined threshold: too many dropped packets imply malintent. However, this heuristic is fundamentally unsound; settingthreshold is, at best, an art and will certainly create unnecessarypositives or mask highly focused attacks. We have designed, develand implemented a compromised router detection protocol that dynaminfers, based on measured traffic rates and buffer sizes, the numbcongestive packet losses that will occur. Once the ambiguity congestion is removed, subsequent packet losses can be attributmalicious actions. We have tested our protocol in Emulab and have stits effectiveness in differentiating attacks from legitimate network beha

8QUIVER: CONSISTENT

OBJECT SHARINGFOR EDGESERVICES

JAVA

We present Quiver, a system that coordinates service proxies placed edge of the Internet to serve distributed clients accessing a sinvolving mutable objects. Quiver enables these proxies to peconsistent accesses to shared objects by migrating the objects to pr

performing operations on those objects. These migrations dramaimprove performance when operations involving an object egeographic locality, since migrating this object into the vicinity of phosting these operations will benefit all such operations. This sreduces the workload in the server. It performs the all operations proxies itself. In this system the operations performed in First-In-Firprocess. This system handles two process serializability and serializabilty for durability in the consistent object sharing . Other workbenefit from Quiver, dispersing the computation load across the proxiesaving the costs of sending operation parameters over the wide areathese are large. Quiver also supports optimizations for single-object that do not involve migrating the object. We detail the protocoimplementing object operations and for accommodating the adinvoluntary disconnection, and voluntary departure of proxies. Finaldiscuss the use of Quiver to build an e-commerce application adistributed network traffic modeling service.


19/32

9

HBA DISTRIBUTEDMETADATAMANAGEMENT FOR

LARGE SCALECLUSTER BASEDSTORAGE SYSTEM

DOTNET

An efficient and distributed scheme for file mapping or file lookup is cin decentralizing metadata management within a group of metservers, here the technique used called HIERARCHICAL BLOOM FARRAYS (HBA) to map filenames to the metadata servers holdingmetadata. The Bloom filter arrays with different levels of accuracieused on each metadata server. The first one with low accuracy and uscapture the destination metadata server information of frequently accfiles. The other array is used to maintain the destination met

information of all files. Simulation results show our HBA design to be effective and efficient in improving the performance and scalability systems in clusters with 1,000 to 10,000 nodes (or super clusters) andthe amount of data in the petabyte scale or higher. HBA is redmetadata operation by using the single metadata architecture insteadmetadata server.

10

DYNAMIC LOADBALANCING INDISTRIBUTEDSYSTEMS IN THEPRESENCE OFDELAYS: AREGENERATION-THEORY APPROACH

JAVA

A regeneration-theory approach is undertaken to analytically characthe average overall completion time in a distributed system. The appconsiders the heterogeneity in the processing rates of the nodes as wthe randomness in the delays imposed by the communication mediumoptimal one-shot load balancing policy is developed and subseqextended to develop an autonomous and distributed load-balancing that can dynamically reallocate incoming external loads at each nodeadaptive and dynamic load balancing policy is implemented and evain a two-node distributed system. The performance of the proposed dyload-balancing policy is compared to that of static policies as w

existing dynamic load-balancing policies by considering the avcompletion time per task and the system processing rate in the preserandom arrivals of the external loads

11

OCGRR: A NEWSCHEDULINGALGORITHM FORDIFFERENTIATEDSERVICESNETWORKS

JAVA

AbstractWe propose a new fair scheduling technique, called O(Output Controlled Grant-based Round Robin), for the support of Diftraffic in a core router. We define a stream to be the same-class pafrom a given immediate upstream router destined to an output port ocore router. At each output port, streams may be isolated in sebuffers before being scheduled in a frame. The sequence of transmission in a frame starts from higher-priority traffic and goes dolower-priority traffic. A frame may have a number of small rounds forclass. Each stream within a class can transmit a number of packets frame based on its available grant, but only one packet per small rthus reducing the intertransmission time from the same streamachieving a smaller jitter and startup latency. The grant can be adjusteway to prevent the starvation of lower priority classes. We also verifdemonstrate the good performance of our scheduler by simulatiocomparison with other algorithms in terms of queuing delay, jitter, and up latency.

12

PFUSION: A P2PARCHITECTURE FORINTERNET-SCALECONTENT-BASEDSEARCH ANDRETRIEVAL

DOTNET

The emerging Peer-to-Peer (P2P) model has become a very powerfuattractive paradigm for developing Internet-scale systems for sresources, including files and documents. The distributed nature of systems, where nodes are typically located across different networkdomains, inherently hinders the efficient retrieval of information. Ipaper, we consider the effects of topologically aware overlay construtechniques on efficient P2P keyword search algorithms. We present theFusion (pFusion) architecture that aims to efficiently inteheterogeneous information that is geographically scattered on pedifferent networks. Our approach builds on work in unstructuredsystems and uses only local knowledge. Our empirical results, usinpFusion middleware architecture and data sets from Akamais Inmapping infrastructure (AKAMAI), the Active Measurement Project (NLand the Text Retrieval Conference (TREC) show that the architectu

propose is both efficient and practical

DEPENDABLE AND SECURE COMPUTING


20/32

1THE EFFECTIVENESS OF

CHECKSUMS FOR

EMBEDDEDNETWORKS

DOTNET

Embedded control networks commonly use checksums to detecttransmission errors. However, design decisions about which checksum tare difficult because of a lack of information about the relative effectivof available options. We study the error detection effectiveness ofollowing commonly used checksum computations for embedded netwexclusive or (XOR), twos complement addition, ones complement adFletcher checksum, Adler checksum, and cyclic redundancy codes (CRstudy of error detection capabilities for random independent bit error

burst errors reveals that XOR, twos complement addition, and checksums are suboptimal for typical application use. Instead, complement addition should be used for applications willing to sacrificedetection effectiveness to reduce compute cost, Fletcher checksuapplications looking for a balance of error detection and compute cosCRCs for applications willing to pay a higher compute cost for fimproved error detection

2

A PRECISE TERMINATIONCONDITION OF THEPROBALASTIC

PACKET MARKINGALGORITHM

JAVA

The probabilistic packet marking (PPM) algorithm is a promising wdiscover the Internet map or an attack graph that the attack patraversed during a distributed denial-of-service attack. However, thealgorithm is not perfect, as its termination condition is not well defined literature. More importantly, without a proper termination conditionattack graph constructed by the PPM algorithm would be wrong. In this we provide a precise termination condition for the PPM algorithm and the new algorithm the Rectified PPM (RPPM) algorithm. The most signmerit of the RPPM algorithm is that when the algorithm terminatealgorithm guarantees that the constructed attack graph is correct, wspecified level of confidence. We carry out simulations on the RPPM algoand show that the RPPM algorithm can guarantee the correctness oconstructed attack graph under 1) different probabilities that a router the attack packets and 2) different structures of the network graph. The algorithm provides an autonomous way for the original PPM algoritdetermine its termination, and it is a promising means of enhancinreliability of the PPM algorithm.

3MODELING & AUTOMATED

CONTAINMENT OF

WORMS

JAVA

Self-propagating codes, called worms, such as Code Red, NimdaSlammer, have drawn significant attention due to their enormously adimpact on the Internet. Thus, there is great interest in the rescommunity in modeling the spread of worms and in providing adedefense mechanisms against them. In this paper, we present a (stochbranching process model for characterizing the propagation of Inworms. The model is developed for uniform scanning worms andextended to preference scanning worms. This model leads todevelopment of an automatic worm containment strategy that preven

spread of a worm beyond its early stage. Specifically, for uniform scaworms, we are able to determine whether the worm spread will evenstop. We then extend our results to contain uniform scanning wormsautomatic worm containment schemes effectively contain both unscanning worms and local preference scanning worms, and it is valithrough simulations and real trace data to be non intrusive.

4

TRUST WORTHY COMUTING UNDERRESOURCECONSTRAINTS WITHTHE DOWN POLICY

DOTNET

In this project we present a simple way to resolve a complicated nesecurity. This is done by the following two ways. They are as follows, fthe decrypt only when necessary (DOWN) policy, which can substaimprove the

ieee projects 2009

Documents