OVERVIEW OF THE PROJECTCloud computing isa model for enabling ubiquitous, convenient, on-demand net-work accessto a shared pool of configurable computing resources that can be rapidlyprovisioned and released with service provider interaction .
It is a new paradigm fordelivering on-demand resources for customers through internet.A service is a mech- anism that is capable of providing one or more functionalities,which it is possibleto use in compliance with provider definedrestrictions and rules and through an in- terface .There are three services models in cloud. Theyare Software as a service: A software or application that is executing on a vendors infrastructure isrecognized as a service providedthat the consumer has limited permission to accessand the provisionis through a thin clientor a program interface for sending data andreceiving results. The consumer isunaware ofthe application providersinfrastructure and has lim- ited authority to configure some settings. Platformas a service: In this servicesmodel, the servicevendor provides moderatebasic requisites, includingthe operating system, network and servers, and development tools to allow the consumer to develop ac-quired applications or software and manage their configurable settings. Infrastructureas a service: The cloud service consumer has developed the required applications and needs only a basic infrastructure. Insuch cases, processors,networks, and storage canbe provided byvendors as services with consumer provisions.
Cloud service Ranking is needed tocloud service consumers to choose appropriate cloud service from a pool of available cloud services. The Qos parameters such as response time, availability, throughput,etc. are used to rank the cloud services based upon consumers require-ments. Particle swarm optimization (PSO) is a computational method that optimizes aproblem by iteratively trying to improve a candidate solutionwith regard to a given measure of quality. PSO optimizes aproblem by having a population of candidate so- lutions, here dubbed particles, and moving these particles around in the search-space according to simple mathematical formulae. 1.2 PROBLEM STATEMENT In cloudservice ranking approach only few quantifiable parameters QoS attributes wereused for ranking.
Several non-quantifiable QoS attributes have major impact inthe ranking and selection process. Also, static ranking of cloud services mayprovide inappropriate cloud service to cloud service consumers as therequirements of one consumer vary with another. The dynamic ranking andselection of cloud services is solved by designing a cloud broker model withseveral components work together to perform Cloud Service Ranking and Selection using Particle Swarm Optimization. 1.3 CHALLENGES AND SCOPE· The accuracy achieved through thisproject is 94% which can be increased further.· The classifiers considered can bechanged further to improve efficiency.· The proposed project is subject to textmining and so still other mining techniques like spatial and correlationtechniques can be used.
CHAPTER2LITERATURE SURVEY2.1 REVIEW In this paper, they survey state-of-the-art Cloud services selection approaches, which are analyzed from the following five perspectives: decision-making techniques; data representation models; parameters and characteristics of Cloud services; contexts, purposes. After comparing and summarizing the approaches they identify the pri- mary research issues in Cloud service selection. Optimization-basedapproaches for Cloud service selection: Chang etal.(2012) developed adynamic programming-basedalgorithm to select Cloud storage providers that can maximize the data survival prob- abilityor the amount of surviving data, subjectto a fixed budget. They formulated the problem of multiple storageservice provider selection into a probability model with clearly defined object functions and cost measurements. The availability of the storage service is quantitatively analyzed bytwo methods minimum failure probability with a given budget, and maximum validitywith a given budget. Sundareswaranetal.
(2012) employed a greedy algorithm- based method for Cloud service selection. They pro- posed the use of the B+ tree to index Cloud service providers (i.e.CSPindexing) and encode services and user requirements. The indexing structure supports the indexing of serviceproperties andthe modeling of their relative importance, as ordered by users. It enables fastinformation retrieval for decision makers. Martensand Teuteberg(2012) developed a scalable mathematical decision model for discrete dynamic opti- mization problems in Cloud service selection.
The model helps organizations to iden- tifysuitable Cloud services by minimizing costs and risks. An AHP-based approach isproposed tomeasure the relative importance of the services in a business process and the relative importance of security parameters in a risk evaluation process. Finally, decisions can be made by solving the formulated mathematical models. Optimiza- tion techniques, such as linear, non-linear, and genetic algorithms, are recommended as the tools for solving models, depending on the specific service outsourcing sce-narios. Identified issues: The open issues on contemporary Cloud service selection approaches are 1. Lack of a marketplacefor Cloud service publication and transac- tion: Cloud services do not have a standard for service publicationand registry. Thelack of detailed service QoS information makes it difficult for service users to make educated purchasingdecisions.
Cloud service allows service users to rate and make comments on services, but there is no feedback from users. 2. Lack of normalizationfor Cloud servicedescription serving different kinds of users: The flourishing of Cloud services highlights the need for a unified specification for Cloud services. A high level of abstraction and support forthe simple publication, discovery, selection, and use of resources for both service providers and users is needed.
3. Lack of a search engine system for the automatic identification and updating of Cloud service information: Cloud service specification lacks a standard form, especially for IaaS and PaaS. The service information is typically published as plain text on a Web page, which usuallynarrows to a functional description rather than being complete enough to include tech- nical details. Such incompleteness prevents keyword-based search engines returning accurate services.
4. Lack of an efficientmeans to deal with qualitative parameters and fuzzy expression:Qualitative non-functional properties such as security and availabil-ity increase the fuzziness of service evaluation. Current techniques are more focused on quantitative criteria that can be measured via precise numerical values such as response time, storage space and network latency. Hence an efficientmethod of han- dling uncertainty and fuzziness in service specification and user requirements needs to be taken into account forthe chosen services. 5.
Lessconcern on multi-tenancy service selection. 6. Lackof an advanced multi-criteria-based measurement ofuser preferences. 7. Lack of consideration of the interdependency ofcriteria 8.Lack of long term performance predication and dynamic application strategy. CHAPTER 3DESIGN AND IMPLEMENTATION3.1 EXISTING SYSTEMThe data mining technique that is beingused comprise of a model that helps in training the train data set.
The modelis made up of techniques without any Cross Validations and repeats. Hence theobtained accuracy is around 92%. The false positive rate is also high. Thoughall kind of vulnerabilities are considered, the results of all vulnerabilitiesare of the same accuracy. The vulnerabilities include XSS, SQL Injection.
3.2 PROPOSED SYSTEM Almostall web applications is moving from a traditional deployment strategy to anon-demand cloud environment. It is highly difficult for the cloud serviceconsumers to choose wisely between theavailable cloud providers. On the other hand, each and every cloud provider mayhave interest on different parameters to be set for their infrastructures.Also, there is no common registry to register the service level agreement ofcloud service as that of the web services.
Hence, it becomes difficult for theconsumers to choose appropriately the required services and thereby cloudservice providers. CHAPTER 4DESIGN AND IMPLEMENTATION4.1 OVERALL DESCRIPTIONThe proposed cloud brokerarchitecture has three components.
Theyare Cloud Service Consumer; theindividual or an organization that requires a cloud service either to deploy anapplication or for application development, Cloud Broker; is the middleware that receives inputfrom the cloud consumers as well as the cloud service providers. It checks theservice level objectives with that of the service level agreement and makes thedecision processing to rank and thereby select the cloud service. Cloud Service Provider; is an entity that providescloud services to the end users or cloud service consumers. 4.2 ARCHITECTURE DIAGRAM The CloudBroker has two databases SLA repository andQos information reposi- tory and has probation manager, rank manager, co-ordinationAgent and search agent. The SLAs of cloud service providersare stored in the SLA Repository of cloud broker.
The SLA document consists of the quantifiable and non- quantifiable Qos parameterswhich include service name, cloud provider, security, availability, processor speed,cost per hour, storage, bandwidth, performance, etc., The Probation Manager : takes SLA from SLA repository and checks the parameters of SLA during the probation period. After the validation it informs the rank managerwith updated parameters.
The Rank Manager: hasrank table and updates therank table with SLA parameter given by probation manager. Rank table contains ranking ofcloud services according to the SLA parameters. If a service is longer used by a consumer, then rank manager gives the service to probationmanager for validation.The Qosinformation repository: feedback of the past customer experienceare storedin Qos Information Repository.
4.3 LIST OF MODULES1. Build SLARepositoryand Design Cloud Broker – Probation Manager- Rank Manager2. Build Qos InformationRepository & add into Cloud Broker- Co-ordination Agent- Search AgentIntegration of cloud service consumers requirements with brokerCloud Service Ranking and Selection using PSO 4.
3.1SLA RepositoryStep1: The SLA from cloud service providers fordifferenet cloud services was col-lected. The SLA document consistsof the quantifiable Qos parametersuca as servicename, Figure 2: Input SLA cloud provider, security, availability, processorspeed, cost per hour, storage and non Qosparameter asbandwidth etc.
The SLA parameters are collected and stored in Mysql server. 4.3.2Design of Cloud BrokerThe cloud brokerhas four entities Probation Manager,Rank Manager ,Co-ordination Agent,Search Agent.
Using cloudsim, the broker is created with entities along with cloudlets. 4.3.3Probation ManagerStep1:Simulation of Probation Manager Figure 5: ProbationManager simulation Step2: TheProbation Managergets the SLAparameters from the database andpopulate the table with SLA. Figure 6: ProbationManager gets SLA’s 4.
3.4 Timeline CHAPTER 5DEVELOPMENT ENVIRONMENT5.1 HARDWARE REQUIREMENTS HARDWARE CONFIGURATION RAM 1 GB and above Processor Dual core and above Hard Disk 80 GB and above Table:4.1hardware requirements 5.2 SOFTWARE REQUIREMENTS SOFTWARE VERSIONS Operating System Windows 7 Application Environment Java(JDK) Programming Language Python Table:4.2software requirements CHAPTER 6CONCLUSION AND FUTURE WORKThedata thus has been filtered to figure out what are the data that are vulnerableand non-vulnerable data. The improved accuracy helps in better filtering ofdata.
The future work is to implement Ensembling models in order to achievestill better accuracy results. Also the method of preventing the vulnerabledata can also be proposed thereby preventing the impact of vulnerable dataduring the transmission of it and safeguarding the entire system.Ensemblingis a general term for combining many classifiers by averaging or voting.
It isa form of meta learning in that it focuses on how to merge results of arbitraryunderlying classifiers. Generally, ensembles of classifiers perform better thansingle classifiers, and the averaging process allows for more granularity ofchoice in the bias-variance tradeoff.Namesof ensemble techniques include bagging, boosting, modelaveraging, and weak learner theory.An obvious strategy isthus to implement as many different solvers as possible and ensemble them alltogether, a sort of “More Models are Better” approach.Text Mining is the keyto determine the vulnerable data at the source and efficient methods inadopting text mining will improve the mining results. CHAPTER 7OUTPUT OF MODULES CHAPTER 8 REFERENCES1. Buyyaet al., ?Cloud Computing and Emerging IT Platforms: Vision, Hype, and Realityfor Delivering Computing as the 5th Utility,Future Generation Computer Systems,vol.
25, no. 6, pp. 599–616, 2009. 2. S.Zhang, C. Zhu, J.
K. O. Sin, and P. K. T.
Mok, ?A novel ultrathin elevatedchannel low-temperature poly-Si TFT,? IEEE Electron Device Lett., vol. 20, pp.569–571, Nov. 1999. 3. N. Thio and S.
Karunasekera, ?Automaticmeasurement of a QoS metric for Web service recommendation,? in ProceedingsAustralian Software Engineering Conference, 2005, pp. 202–211. 4. J.Marden, Analyzing and Modeling Ranking Data. Chapman & Hall, 1995. 5.
P.A.Bonatti and P. Festa, ?On Optimal Service Selection,? Proc.
14th Int’l Conf.World Wide Web (WWW ’05), pp. 530-538, 2005. 6. J.S.Breese, D. Heckerman, and C.
Kadie, ?Empirical Analysis of PredictiveAlgorithms for Collaborative Filtering,? Proc. 14th Ann. Conf. Uncertainty inArtificial Intelligence (UAI ’98), pp. 43-52, 1998.
