IMPERSONATION DETECTION IN DATA CAPTURING SYSTEMS (DCS) WITH IMAGE DATA ANALYSIS USING MACHINE LEARNING

Abstract

Impersonation attacks pose a significant threat to the security and integrity of data-capturing systems (DCS). With the increasing reliance on image data in various domains, it becomes crucial to develop effective methods for detecting impersonation attacks using image data analysis. This research aims to address the problem of impersonation detection in DCS by leveraging machine learning techniques. Specifically, this thesis proposes a novel approach that combines image processing, feature extraction, and machine learning algorithms to identify and detect impersonation attacks in DCS. Many of today's applications require cooperating with multiple database systems. The heterogeneous data representations of databases include schemas, actual data etc. The data inconsistencies in databases may occur for semantic inconsistencies stored in syntactical data; the data is distinguishable but semantically identical or for the same record representation. Semantic heterogeneity in database systems has a traditional problem that always results in data duplication i.e., the same record is accumulated two or more times in multiple database systems. The problem of Data Duplication is that its highly pervasive in legacy software systems. Data duplication refers to a data source having more than one record for the same identity, mostly with different syntaxes for the same item/identity. This problem has been known as highly important to many entities due to the size and complexity of today's database systems. Several researchers tried to resolve the data duplicity problem by using different techniques such as the Sorted Neighbors Method (SNM), identity matrix, and many clustering methods. However, an Image classification linked approach has not been used for these studies. In recent studies, Performance comparison on image classification with the accuracy of the ML models, fuzzy measure, decision tree, artificial neural network, as well as other support vector machine methods oriented results are obtained from the literature survey.