International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 07 | July 2022
p-ISSN: 2395-0072
www.irjet.net
Analysis on Fraud Detection Mechanisms Using Machine Learning Techniques Anusree Sanalkumar Computer Science and Engineering Department, Toms College Of Engineering, Kottayam, Kerala, India ---------------------------------------------------------------------***--------------------------------------------------------------------1.1 OBJECTIVE Abstract - A blockchain is basically a decentralized digital ledger of transactions that is duplicated and distributed across the complex network systems. Each transactional data is stored as a block in the network. Every blockchain transaction is created across a peer to peer network and is authenticated by the digital signature of the owner. Hence, the information that contained in the ledger is highly secure. Even though they are secure, fraudulent activities still takes place. Many machine learning techniques are used for the reduction of fraudulent activities and its detection. In this paper various machine learning techniques are studied and combined to produce a more efficient fraud detection mechanism by Ensembling the most prominent machine learning algorithms. Algorithms like Random Forest Classifier and Adaboost is ensembled using the Stacking method and that accuracy is measured and compared with their individual accuracy results.
Different types of machine learning techniques have been used to solve the problem of online fraudulent activities. There are supervised and unsupervised machine learning techniques used in different methods to detect fraudulent activities in the online transaction system. The blockchain technology also uses this machine learning techniques as it is the most reliable and fast method in the present time. Different prominent machine learning algorithms are ensemble together to provide a better fraud detection mechanism and this is analyzed with the existing system to acquire the result of finding which method is better for the detection mechanism in machine learning techniques.
2. LITERATURE SURVEY Madhuparna Bhowmik et al, [1] A method has been proposed for the detection of fraudulent transactions in a blockchain network using machine learning.they evaluate all our classification models using bootstrap sampling and processed the data using node2vec algorithm.
Key Words: Blockchain, fraud detection, machine learning, Ensembling, online transaction
1. INTRODUCTION
Yuanfeng Cai et al. [2] discussed the objective and subjective frauds. They conclude that blockchain effectively detects objective fraud but not subjective fraud and thus uses Machine Learning to mitigate the weakness.
Machines are a great asset at processing and validating large datasets. They are able to detect and recognize thousands of patterns on a user’s browsing in networks and their transactions. We can predict fraud in a large volume of transactions by applying several computing technologies to raw data. This is one of the reasons why we use machine learning algorithms for preventing fraud for our clients.
Jennifer J. Xu [3] discussed the types of fraudulent activities that blockchain can detect and the ones that blockchain is still vulnerable to. This paved a path towards ideas about what problems a Machine learning part needs to consider. She specifies that attacks like Identity theft and system hacking are still possible and challenging to detect using blockchain as it just uses some predetermined rules.
Fraud detection process using machine learning starts with gathering and preprocessing the data. Then various machine learning models is fed with training sets to predict the probability of fraud. It is a very useful technology which allows us to find patterns of an anomaly in everyday transactions. Machine learning technologies has become a superior mechanism in finding and preventing these fraudulent activities in the network system
Michał Ostapowicz et al. [4] used Supervised Machine learning methods to detect fraudulent activities. They focused on the fact that malicious actors can steal money by applying well-known malware software or fake emails. Therefore they used the capabilities of Random Forests, Support Vector Machines, and XGBoost classifiers to identify such accounts based on a dataset of more than 300 thousand accounts.
Different prominent machine learning algorithms like SVM, Adaboost, Random forest classifier etc. are ensemble together to provide a better fraud detection mechanism and this is analyzed with the existing system to acquire the result of finding which method is better for the detection mechanism in machine learning techniques.
© 2022, IRJET
|
Impact Factor value: 7.529
Blaˇz Podgorelec et al. [5] devised a method using Machine Learning for the automated signing of transactions in the blockchain. Hence, it also uses a personalized identification of anomalous transactions.
|
ISO 9001:2008 Certified Journal
|
Page 629