This project entitled “Credit/Debit Card Fraud Detection System Using Hidden Markov Model” submitted by S. Asifur Rahman and Nusrath Jhahan to Department of Computer Science and Engineering, Daffodil International University has been accepted as satisfactory for partial fulfillment of the requirement for a B.Sc. Department of Computer Science and Engineering Faculty of Science and Information Technology Daffodil International University.
We declare that this project will be carried out by us under the supervision of Mme. We also declare that neither this project nor any part of this project has been submitted elsewhere for the award of any degree or diploma. Her endless patience, scientific guidance, constant encouragement, constant and energetic supervision, constructive criticism, valuable advice, reading many poor drafts and correcting them at every stage made this project possible to complete.
Syed Akhter Hossain, Professor and Head of CSE Department for his kind help in completing our project and other faculty members and staff of CSE Department of Daffodil International University. We would like to thank our entire course staff at Daffodil International University who participated in this discussion while completing the course.
Introduction
Motivation
Objective
Advantages
Report Layout
Introduction
Initially, the fraudster takes control of the account by providing the customer's account number or credit card number. The fraudster will then adjust the details of the card so that they fully match the details of the original cards. Here I would like to briefly discuss some of the credit card fraud detection techniques given in the past.
In these four steps, the system analyzes the spending behavior of past and present customers together based on a specific purchase. The high-level use case diagram is shown in Figure 3.1 and is discussed below in terms of the functionalities and the relationships between different use cases. In this section, we describe some use cases of the use case diagram.
From the initial probability given to the system, it calculates the probability of the next transaction based on some hidden states. So the system will record the next genuine transaction if the result is 50% or more. If it is less than 50%, the system will ask the user some security questions that are only known to the original user.
Requirements Analysis: This is the first stage of this methodology where we first need to identify the problem to solve it and then find out all the possible requirements of the system to develop the system. Here the user tests the system to ensure that the functional requirements defined in the requirements document are met by the developed or modified system. After completing the construction phase, the system is well acquainted with the observation symbols and their sequences, namely O1, O2 ..ON.
So, when a new transaction occurs, the system will add it to the watchdog ON+1 and delete the first transaction. If any unexpected probabilistic deviation occurs during the transaction, the system will detect it as a fraudulent transaction. In my proposed system I have shown that the system first checks the probability that the next transaction is real or a fraud.
So, when the probability of the next transaction based on some states is less than 50%, the system will detect it as a fraudulent transaction. Because if the system takes more time to analyze the user's transaction patterns, in turn the users will be safe less time. It is important to note that evaluation is the main objective of this project.
The system will calculate the probability that the next transaction is fraud or real based on some facts.
Comparative Study
Related Work
The accuracy rate of the soft system is very high and it also has a low false alarm rate. One is for incoming transactions in the past and the other is for genuine transactions made in the past. However, it does not have the ability to detect duplicate transactions or clone credit card fraud.
Generally, it uses artificial intelligence programming and machine learning methods in neural network and Bayesian network is used for developing specific pattern to understand the spending behavior of customers.
Hidden Markov Model
Challenges
REQUIREMENT SPECIFICATION………………. 08-14
Use Case Model
- Use Case Description
Description: This use case describes the scenario where the user can know the details of his/her user account. Description: This use case describes the scenario where the user can withdraw their deposited money.
Complete System Diagram
If he/she is able to answer these questions, then the system will enable the transaction procedure.
Design Requirements
- Home Screen Design
- Login Screen
- Account Details Screen
- Transaction History Screen
- ATM Screen
When a user starts or opens our app, they can see this home screen first.
Methodology
Literature Study: At this stage we had to study about the project work, what is the process to solve it, the best way to solve it. This phase helps us to make a decision about the right system to develop a project/system. This system design helps specify the hardware and helps define the overall system architecture.
Here we thought about how the program implements or executes, which process should be implemented first, etc. In short, in this phase we planned all kinds of implementation processes to develop our system. Implementation: The implementation plan is given in the previous phase and the detailed specification prepared during the design phase is translated into hardware, communication and executable software.
Each device is developed and tested for its functionality, which we called unit testing. Testing: All devices developed in the implementation phase are integrated into a system after testing each device. Sometimes there are some problems or problems appear in the user environment, to solve these problems maintenance is very important.
Interaction Design and UX
Implementation Requirements
From the following description it is known that an HMM model needs two parameters N and M. In this chapter we will discuss the implementation, evolution criteria and testing part of our project.
Implementation
This is an internal process and it will not affect the online transaction of other users.
Working with random data
- Discussion on the result
It is much faster and simple to compare with the others and also reduces the probability of fraudulent transactions in huge amounts. For the proposed system, I have used several data sets which are generated by a random number generator. It is possible to give a more correct answer if the analysis on the transaction increases to more than 10.
Sometimes it may happen that the user is genuine, but the probability of that transaction is below the threshold value. As we gain a better understanding of algorithms and their implementations, we are better prepared to intelligently design our estimation method. We will consider this project successful when we are sure that we have properly collected all the information and helped people.
In this project I have discussed the common ways of credit card fraud and I have used an HMM based system which will help to detect credit card fraud transactions. To sort the facts I have created three categories low, medium and high for the users. The future work for this system may be to find out more common human behavior to make this system more secure.
We will combine this system with the neural network so that it will work more efficiently.
Evolution
Testing
CONCLUSION AND FUTURE SCOPE
Conclusion
Limitation
Future Scope