ONLINE SELLING

CHAPTER ONE

INRODUCTION

1.1            Background of the Study

Online Selling sites such as eBay2 and Trade Me3 allow goods and services to be bought and sold online anonymously. The most common type of online Selling is the English Selling, where bids are placed in ascending order, publicly observable, and the winner is the final bidder with the highest bid (Menezes et al., 2015). In 2013, eBay had 128 million active users and selling volume of more than $22 billion USD (Hainline, 2014). The volume of auctions, user anonymity, and low barriers to entry make online selling a lucrative target for fraud. The anonymity and simplicity of creating multiple aliases allow unsuspecting users to be exploited by dishonest users. This exploitation can take many forms, including shilling, non-delivery, misrepresentation, or the sale of stolen goods (Dong et al., 2009).

Dishonest users will also disguise themselves to avoid detection by imitating normal behaviours (Chang et al., 2012), making fraudulent behaviours difficult to define. Previous work has noted that users often appear to behave irrationally (Mizuta et al., 2011), and previous attempts at clustering users into predefined types according to their bidding behaviour have failed to label the majority of users (Shah et al., 2013). The range of potential fraudulent behaviour together with the number and range of legitimate behaviours makes it difficult to differentiate between fraudulent and legitimate users.

Frauds committed before or after Selling, such as misrepresentation or non-delivery fraud, are generally identified after-the-fact, for example, when the buyer never receives the item they paid for. Detection of these frauds generally relies on real-world evidence as opposed to online prevention and detection mechanisms (Dong et al., 2009). This contrasts with in-auction fraud, such as reputation fraud or shilling, which occur while the Selling is in progress. It can often be difficult to determine when and if in-auction fraud has occurred.

The majority of existing research use machine learning to detect or predict in- Selling fraud. A range of methods have been used, including various regression models such as logistic regression and probity models; graphical models such as neural networks, Markov random fields and decision trees; and rule-based systems of the proposed methods, the majority are supervised, which has several drawbacks.

The E-commerce proportion in global retail expenditure has been steadily increasing over the years showing an obvious shift from brick and mortar to retail clicks. To analyze the exact problem of building an interactive models for the identification of auction fraud in the entry of data into e- commerce. This is why the most popular site's business develops with retailers and other auction customers. Where viral customers purchase products from online trading, customers may worry about fraudulent actions to get unlawful benefits from honest parties. Proactive modesty systems for detecting fraud are thus a necessary practice to prevent such illegal activities. The shopping product is built according to the customer's requirements and is safer online and resting, and the rules and regulations that are necessary to follow no longer seem to be the best of workable selection, coefficient limits that facilitate the shopping product and make it easier for the user model to compete on each platform so that it can experiment. With the advancement of cutting-edge technology and global connectivity, fraud has risen dramatically. There are two ways to identify fraud: prevention and detection. By serving as a layer of defense, prevention helps to keep fraudsters at bay.

 

 

 

1.2     Statement of the Problem

Online Selling are inherently insecure due to geographical scale and lack of accountability. Since participants are not physically present at the Selling proceedings there are opportunities to cheat. For example, bidders can repudiate or forge bid and sellers might not deliver goods. There are two main difficulties with producing effective online auction fraud detection algorithms: first, the lack of publicly available real data, and second, the nature of online Selling fraud. Commercial companies who have access to this data are reluctant to publicly release it for many reasons, including privacy concerns, and the potential for loss in user confidence if a high number of fraudulent activities are identified. The nature of online auction fraud also makes it difficult to develop effective fraud detection algorithms, supervised or unsupervised. In many fraud detection domains, normal cases vastly outnumber fraudulent ones.

The proposed system will help to tackle the problems stated above, whereby a software system will be developed to detect shilling fraud, reputation fraud, non-delivery fraud and misrepresentation fraud effectively.

1.3     Aim and Objectives of the Study

The aim of the study is to develop a system that will be able to detect selling fraud, reputation, fraud, shielding fraud and misrepresentation fraud in an online Selling.

Objectives of the Research Are As Follow:

i.        Provide a security model for auctioning online that protects bidders’ personal information.

ii.       Develop methods to detect certain types of auction fraud (i.e. shilling).

iii.      Explore the security implications of agent based negotiation.

iv.      Construct a model for securely and anonymously trading shares online.

v.       Devise and evaluate alternate software mechanisms for clearing Continuous Double Auctions (CDAs).

1.4     Significance of the Study

This study serves as a contribution towards improving information security and fraud detection in an online auction. Online auction fraud detection will be of immense benefit because:

i.        It allows an auctioneer to set up an auction and advertise it i.e., the type of goods being auctioned, starting time, etc.

ii.       It allow participant to register for the auction and be giving bidding keys for anonymity.

iii.      It allows registered bidder to submit their bids and views auctions result when the auction time is over.

iv.      It automatically detects fraud when a bidder tries to defraud an auctioneer.

v.       Once a bidder repudiate after winning an auction, the auctioneer can reveal the bidders identity.

vi.      Malicious bidders can be easily revoked from all future auctions

1.5     Scope of the Study

The essence of this research work is to primarily develop a system that will detect fraud in an online auction. The research work will cover the development of an algorithm that will be able to automatically detect shilling fraud, reputation fraud, shielding fraud and misrepresentation fraud. The software will also cover the creation of a bulletin board where all auctions will be published for users to and rate the auctioneers and bidders.

 

1.6     Definition of Terms

i.    Auction: An auction is a process of buying and selling goods or services by offering them up forbid, taking bids, and then selling the item to the highest bidder.

ii.   Bidding: Bidding is an offer (often competitive) to set a price by an individual or business for a product or service or a demand that something be done. Bidding is used to determine the cost or value of something.

iii.  Fraud: Is defined as deception deliberately practiced with a view of gaining an unlawful or unfair advantage.

iv.  Identity theft: n this is an action that proceed enable a fraud to occur.

v.   Internet fraud: this refers generally to any type of fraud scheme that uses one or more components of the internet such as chartroom, e-mail, message boards or websites to present fraudulent transactions or to transmit the proceed of fraud into in financial institutions or to other connected with the scheme.

vi.  Algorithm: Is defined as a description of a procedure which terminates with a result.

vii.Shielding: An illegitimate way to preserve a low bid in an online auction. It takes three people. The first places-a low bid and the other two immediately bid high and keep bidding higher, which is intended to eliminate all other interested parties. At the last minute, the two high bidders drop out, and the low bidder wins by default.

viii.Repudiate: a process whereby a bidder refuse to acknowledge making a bid after winning an auction.

ix. Anonymous bidding: A bidding structure in the market in which the identities of the bidder behind a bid is masked, which is used to prevent any advantages from this knowledge such as discriminating against specific buyers or sellers.

x. Model: A learning model can be a mathematical representation of a real- world process. To machine generate a machine learning model you will need to provide training data to a machine learning algorithm to learn from.

 

 

CHAPTER TWO

LITERATURE REVIEW

2.1 Theoretical Framework

Selling is Latin word which means augment. Selling, have existed for centuries.

Selling is a bid, a process of selling or buying and services offered take place.

I There are several" different. Types of selling and certain rules exist for each

Selling. There are variations for selling which may include minimum price limit, maximum price limit and time limitations etc. Depending upon the auction method bidder can participate remotely or in person. Remote auction include participating through telephone, mail, and internet. Many economic transactions are conducted through selling. Governments sell treasury, bills, foreign exchange, publicly owned companies, mineral rights, and more recently airwave spectrum rights via auctions. Artwork, antiques, cars, and houses are also sold by selling.

Government contracts are awarded by procurement auctions, which are also used by firm to buy inputs or to subcontract work (Noufidali et al, 2013).

A shill is a person who pretends to be a legitimate buyer and feigns enthusiasm for an Selling item by bidding up the selling price. The role of a shill is typically played by an associate of the seller. In some cases, it can also be played by the Seller himself, who poses as a legitimate buyer under a fake online user ID. The ultimate purpose of employing shills is to trick legitimate buyers into paying more than they would if there were no selling frauds (Dong et al., 2009).

While shilling is recognized as a problem (Bhargava. et .al., 2005), established means of defence against shills did not work in live auctions and did not focus on some important shill bidding behaviours (Chau et al., 2009; Dong et al., 2009), for example, multipl-e consecutive bidding by the same user, bidding with different identities. The advent of online selling such as eBay, Amazon and bid has made shill bidding much more exploitable. This is because it is relatively simple for a seller to register under many aliases and operate in rings with impunity (Read etal., 2006). Furthermore, as bidders are not physically present it becomes much easier for a shill to anonymously influence the bidding process.

2.2 Empirical Review of Precious Work in the Area of Study

Feedback Based Reputation Systems

Feedback based reputation system is a one of the simplest fraud detection system.

In these systems help buyers decide whether to purchase a product based on a feedback score. After the completion of selling, both the seller and the buyer can put down ratings and feedback comments on the other party. These comments and

Feedback build up in the trader's transaction history, of which the feedback score is one Part. So this type of reputation system' is simple and easy to understand. It uses

Positive, neutral, and negative to denote the level of fulfilment for a trade. Analyse the reputation system capacity of finding fraudulent behaviour; it has major drawbacks (Chau et al., 2009). Most online selling houses agree to passive approaches to the coordination of reputation systems and management policies that could address fraudulent schemes. Though, if users had more proactive approaches, such as an automatic fraud detector, online auctioning could be safer (Noufidalietai, 2013). The reputation systems development, the most selling houses are used simple method. The auction houses used the method of marking

I rates, after a transaction is completed, the traders can only be rated on a positive, neutral and negative scale. The both buyer and seller want to be a good name, so they try to increase the positive feedback, so some of them try to increase their reputation with ambiguous means. Crook could attract buyers by fabricating transaction records that inflate their feedback scores in order to hide their malicious intent (Read et al. 2006). In online selling the buyer using the common fraudulent trick is to first make several small businesses in order earn more positive feedback rating score, but then cheat later on the first expensive product.

Another common type fraud is using multiple identities in online selling. In this type a fraudster first creates multiple identities, dividing them into two groups, fraudsters and legitimate. Then, the fraud stars use the legitimate to artificially boost their reputations by leaving positive ratings (Tatenda et al, 2016).

Data Mining Based Fraud Detection: the other name for data mining is knowledge discovery, it is a powerful computer-assisted process designed to analyse and take out useful information from historical data (Michel, 2010). It allows users to analyse data from different magnitude or perspectives in order to uncover consistent patterns, anomalies and systematic correlations between data elements. The data mining technique is used in different areas, the ultimate objective of data mining is to predict future behaviours and trends based on the discovered patterns and association rules. In the area of research most of the researchers have adopted data mining methods to detect shill associations and suspicious patterns (Pandit et al, 2007). The main steps for implementing this approach for fraud' detection within a business organization are as follows (Shai et

al., 2013).

i. Data collection and understanding. .,

ii. Data cleaning and preparation for the algorithms.

iii. Experiment design.

iv. Evaluation results in order to review the process.

Data mining approaches, like reputation approaches, also require analysing huge amounts of historical data, and therefore take a very long time to get results. Using data mining approaches do have the advantage of accuracy compared to reputation systems (Tatenda et al, 2016).

2.3 Conceptual Framework

Although the number of sellers and buyers attracted by online selling is growing rapidly. This contemporary business medium faces an important challenge – selling fraud (Albert, 2002). Both sellers and buyers can participate in selling fraud for their own benefit. Data released by the U.S. Federal Bureau of Investigation's Internet Crime Complaint Centre (IC3) reveals that 93,771 selling complaints were received in 2006, representing 45 percent of all Internet fraud complaints. (McMillan 2008). Selling complaints remain the largest source of Internet-related complaints, consistently ranking at the top of the list for many years (Albert,

2002).

According to Internet Crime Complaint Centre (IC3), there are several ways online selling fraud can occur: misrepresentation of a product for sale, non-delivery of merchandise or services sold, triangulation (fraudsters purchase items using a stolen credit card, selling the items to uninitiated buyers thereby retaining the cash and transferring the risk of seizure to the end recipient), fee stacking (charging extra money after an auction is over), selling black market goods, multiple bidding (buyers inflate prices using aliases, which frustrates competitors, then at the last moment the high bids are withdrawn to secure a low bid), and finally shill bidding (sellers or their associates place bids on their own selling for fraudulent purpose Shaietai, 2013).

To understand online selling fraud, it is convenient to first classify the various types of online selling fraud according to the three time periods in which the fraudulent behaviour can take place: pre-selling, in-selling and post-selling.

Misrepresentation of items, selling of black market goods and triangulation usually

Occur before the selling start, so we classify them as pre-selling fraud; and non- delivery of goods and fee stacking occur after selling close, so we consider them as post-selling fraud. In-selling fraud is the main focus of this paper. In-selling fraud happens while transactions are in progress, thus it may occur without leaving direct physical evidence, and worst of all may not even be noticed by the victims.

In addition, while pre-selling fraud and post-selling fraud have already attracted researchers' and policy makers' attention, in-selling fraud has attracted much less attention due to its complexity in detection (Ba et al., 2012). In order to reduce the loss to victims and to protect, online business participants, in-selling fraud deserves more attention and effort from mechanism designers and information technology researchers (Grazioli., 2014).

Many consumer guidance websites and newspapers have provided selling fraud detection tips, such as checking if a seller and a suspicious bidder are from the same geographic region; and searching a shill suspect's -bidding history to  determine if a seller and the shill suspect have a partnership. However, even if a suspected shill and a seller are located far from each other, they still could be partners (Pandit et al, 2007). With modern Internet communication applications, sellers and shills can communicate with each other very easily as if they were sitting next to each other. In addition, whereas there may be a large number of historical records for long-time sellers, "it is difficult and time-consuming for a bidder to discover the partnership between sellers and shills. Unfortunately, in- selling fraud is so sophisticated and tricky that such tips are extremely difficult for consumers to effectively apply (Albert, 2002).

Many researchers from economics, business, system science and computer science have realized the severity of this problem (Grazioli, 2014). They are working to combat selling fraud and have produced some preliminary work. In this paper, we aim to provide an overview of the state-of-the-art Internet in-auction fraud prediction, prevention and detection techniques, and to highlight challenging research issues, in "this interesting new area. The surveyed countermeasures are from both the economics and the computer science perspectives.

2.4 Review of Related Empirical Studies

Xu (2008) presented a Multi-Agent Trust Management (ATM) framework for online selling. The shill inference procedure was embedded in the security agent of ATM. (Xu, 2008) introduced a formal model checking approach to detect shilling behaviours, especially the competitive shilling behaviours (Cheng, 2007).

Wood statistically analysed data from rare coin auctions on eBay, and empirically tested the questionable bidding behaviours that are attributable to shill bidding

(Wood, 2003). Read (2009) designed an algorithm to detect collusive shill bidding where multiple shill bidders shill in a group. But the two problems of duplicate identity shill bidding and consecutive multiple bidding are not addressed. Moreover, the algorithm does not work in live selling. (Chau et al., 2009) applied data running and mist propagation-techniques to detect fraudulent users in online selling systems.

Generally, these techniques suffer from two drawbacks. Data mining related approaches need to deal with a large amount of historical data; thus they may have limited value in -detecting shill bidding in a time-efficient manner. Pattern matching based and model-checking based approaches do not regularly update prior knowledge with the presence of new evidence. Therefore they may frequently generate false positive results. This study proposes an approach in which we detect suspicious shilling behaviours efficiently, and can also make the results more accurate for online selling by updating the training set on the presence of new evidence (Chau et al., 2009).

Dempster-Shafcr theory: Dempster-Shafer (D-S) theory, a probabilistic reasoning technique, is designed to deal with uncertainty and incompleteness of available information. (Dong et al. 2009) proposed a formal approach to verifying ghill bidders using D-S theory (Shafer, 1976). The verification approach utilizes additional evidence, such as various bidding histories and, statistics regarding bidder and seller interactions, to verify if an online bidder is a shill. The belief of whether a bidder is a shill is calculated using the D-S theory, which allows the verifier to reason under uncertainty. If the belief of a bidder for being a shill exceeds a certain threshold, the bidder is marked as a shill bidder. These techniques, however, suffer from being time consuming in their investigation of bidders. Since most bidders do not behave suspiciously

Multi-state Bayesian network: This is a probabilistic graph model that can be used to capture uncertain knowledge in a natural and efficient way. Goel (2010) used a multi-state Bayesian network to verify detected shill suspects. Similar to the

D-S theory-based approach, Bayesian networks are capable of reasoning under uncertainty and can be used to calculate the probability of a bidder being a shill.

This technique also suffers from being time consuming in their investigation of bidders.

Net Probe: This uses belief propagation over Markov Random Fields to classify users in online auctions as honest, fraud and accomplices (Shashank, 2007).

However, Net Probe misclassify nodes in cases where it flips an honest user to an accomplice or where fraudulent users might easily exploit its assumptions to camouflage themselves- The Net Probe algorithm works under the assumption that fraudsters are connected to accomplices with high probability (0.9) and it connects to other fraudsters or honest people with a very low probability (each having a probability of (0.05). Also accomplices are connected to accomplices with a very low probability (0.1) (Shashank, 2007). Hence the Net Probe algorithm essentially assumes a bipartite graph where fraudsters are disconnected from other fraudsters and honest people while accomplices are disconnected from accomplices. Unfortunately these probabilities of connection are not something empirical but an educated guess and might misclassify users.

2.5 Summary of Literature Review

Analysis of the current trend of internet usage shows an exponential proliferation in the online selling sites and their users. An equal and opposite downside to this online practice is the fraud happening alongside. All these transgression remained to be veiled till recent years and now the real time reports highlight stumbling numbers of online frauds.

For the above reason, this study presented an algorithm to detect fraud in e selling. Although many selling websites have taken some selling to avoid shill bidding, there still take place a lot of shill bidding cases from time to time. To find more efficient approach to discourage shilling is therefore of great value.

To eliminate shilling problem, this study proposes a shilling detection algorithm integrated with an online selling.

 

 

 

 

 

 

 

CHAPTER THREE

RESEARCH METHODOLOGY

3.1 Methodology

The methodology used for this research is the waterfall method. The water fall method clarifies the software development process in a linear sequential flow that any phase in the development process begin only if the earlier phase is completed. The development approach does not define the process to go back to the previous phase to handle changes in requirement. Waterfall model is simple and easy to understand. Moreover, each phase has specific deliverables and individual review process. The water fall model uses the following processes: Requirement analysis and definition, System and software design, Implementation and unit testing, Integration and system testing, Operation and maintenance.

3.2 Method of Data Collection

Data collection   is   an important aspect of any type of research study. Inaccurate data collection can impact the results of a study and ultimately lead to invalid results. Data collection methods for impact evaluation vary along a continuum. At the one end of this continuum are quantitative methods (which when viewed in terms of source of data collection refers to the primary source of data collection and at the other end of the field are qualitative methods (likewise, this is known at the secondary source) for data collection.

For this study, both defined methods were used for obtaining data. From the quantitative method, observation is the major avenue which help the researcher to state the problem statements and defines its objectives. While the qualitative methods (secondary source); resources for this study were gathered from journals, textbooks, library on orphanage information system works and the internet too was very helpful. All extracts made from the sources ‘of data are duly cited and referenced.

3.3 Analysis of the Existing System

There are two main difficulties with producing effective online auction fraud detection algorithms: first, the lack of publicly available real data, and second, the nature of online auction fraud. The nature of online auction fraud makes it difficult to develop effective fraud detection algorithms, supervised or unsupervised. Firstly, in many fraud detection domains, normal cases vastly outnumber fraudulent ones. Secondly, many different types of fraud exist ‘in online auctions, each of which can be committed using different strategies. It is difficult to develop algorithms that can detect these diverse -types of fraud, especially when normal behaviour of users is difficult to model as well. Thirdly, as fraudulent accounts are identified and removed, remaining accounts and any new accounts may change their behaviour to avoid detection.

 

3.3.1 Advantages of Existing System

       I.            Bidder can participate in an online auction from anywhere and at any time.

    II.            Bidder can bid at the comfort of their house.

 III.            Bidder can compare prices and decide which one is affordable for their budget

 iv.      Users can get a high quality product at a low rate

3.3.2 Disadvantages of the Existing System

i. Bidders are anonymous

ii. Lack of product genuineness

iii. Some auction websites are fake

iv. Lack of security

3.4 Analysis of the proposed System

Current and emerging state-of-the-art auction platforms rely heavily on information technology, and ‘hence any weakness in information system could be utilized by malicious users to maximize their own profits. In order to overcome this wearies, an online auction fraud detection site is designed as a supplement where users can create accounts, post auctions and place bids as well. A database was designed to store auction data in discretized format. This was done to enable data collection and analysis for each bidder/user in real time. Data collected from each auction was scrutinized to determine each user's bidding behaviour whether it is suspicious or normal. This is achieved using an efficient shill detection Algorithm that can successfully detect the presence of selling fraud using the analysed data immediately after it happens. The system is also able to measure an Auctioneer reputation which will give bidders confidence to conduct a selling with the auctioneer based on the ratings giving to the auctioneer.

The development of the proposed system is very necessary because from the analysis of the existing system it is obvious that its working functionalities are not well integrated. The proposed system will lessen the problems encountered by the present system.

3.5 System Design

System design which is the third stage of systems. Development deals with the

Implementation of the analysis of the system. In carrying out the design, some

Issues were considered such as; the human-computer interaction that makes it possible for the communication between ‘the computer system and the user, the database design, input/output design and specification and the general information/data security. The details of the software specification and design are shown in this chapter. The database and the sub systems that make up the program design are explained.

 

3.5.1 Program Flowchart

The program flowchart depicts the step by step sequence of instructions, work flow or process, showing the steps as boxes of various kinds, and their order by connecting them with arrows. The diagrammatic representation illustrates a solution model to the existing system indicating the start operation and series of actives that occurred before the termination of the system processes by the stop operation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.5 Flow Chart Diagram of the proposed system.

3.6 High Level Model of the Proposed System

Hierarchy diagram shows hierarchical relationships progressing from top tobottom. Hierarchy diagrams are often used to represent the business and corporate. Usually the hierarchy diagram start with a top node and continue with a tree until reaching the lower levels within an organization.

 

 

 

 

 

 

 

 

 

 

 

 

Fig. 3.6       Hierarchical diagram of the proposed System.

3.7 Unified Modelling Language

Unified Modelling Language (UML) is a diagrammatic object-oriented modelling language. It also provides several diagrams such as Use case diagram, Activity diagram, etc. It uses diagram to document an object based decomposition of system showing the interaction between these objects and the dynamics of these objects. UML aims to provide a common vocabulary of object based terms and diagramming techniques that are rich enough to model any system development project from analysis to design. This System is modelled with:

       I.            Use case diagram

    II.            Activity diagram

 III.            Class diagram

3.7.1 Use Case Diagram

The use case diagram of the new system illustrates the users of the system and

Various actions that each of these user can perform.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.7.1 Use Case Diagram of the system

3.7.4 Activity Diagram

This illustrates the dynamic nature of the proposed system by modelling the flow of control from activity to activity.

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.7.2: Activity Diagram of the System

3.7.3    Class Diagram

The purpose class diagram is to depict the classes within a model. In object oriented application, classes have attributes, operations (methods) and relationship with other classes.

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.7.3: Class Diagram of the system

3.8 Database Design

To produce a detail data model of any database, a process involving the design of interface either logically or physically must come up. Where the logical data model contains all the needed physical design choice and physical parameters needed to generate a design in a data definition language, which can then be used to create a database. A fully attributed data model contains details attributes for each entity.

Therefore, the design of the database will be logical design of the database structure use to store data. However, in a relational database these representations are converted to tables.

Table 3.1 Bidder Table

S/N	Field Name	Type	Description
1	Userld	Varchar(15)	Bidder User Identification
2	First Name	Varchar(20)	Bidder First Name
3	Last Name	Varchar(20)	Bidder Last Name
4	Address	Varchar(50)	Bidder Address
5	Gender	Varchar(10)	Bidder Gender
6	Password	Varchar(40)	Bidder Login Password
7	Email	Varchar(20)	Bidder Email Address

 

Table 3.2 Selling Table

S/N	Field Name	Type	Description
1	Sellingld	Int(5)	Selling id
2	StartPrice	Vachar (20)	Selling Starting Price
3	Description	Vachar (20)	Description of Selling
4	StartTime	Vachar (20)	Time of starting  selling
5	Expiration	Int(10)	Expiring time for the selling
6	Status	Int(3)	Selling status i.e progress or ended