Email Fraud Detection

EasyChair Preprint 10523

Abstract

With the rise of IoT, there has been an increase in spamming problems on social media platforms and applications. Researchers have proposed various spam detection methods to address the issue. Despite the existence of anti-spam tools and techniques, spam rates remain high, particularly with the prevalence of malicious emails that contain links to harmful websites. Spam emails can slow down servers by consuming memory or capacity. Filtering emails is one of the most essential approaches to detecting and preventing spam, and various tools for deep learning and machine learning, including Naive Bayes, decision trees, SVM, and random forest, have been employed to achieve this goal. This study classifies the various machine learning approaches used for spam filtering in email and IoT platforms. In addition, the problem of SMS spam messages is increasing globally as the number of mobile users increases, combined with the low cost of SMS services. To address this issue, this paper proposes employing a suite of machine learning techniques to identify and eliminate spam. The experimental results showed that the TF-IDF with Random Forest classification algorithm achieved the highest percentage of accuracy compared to the other algorithms tested. Since the dataset is unbalanced, it is not possible to evaluate performance based just on accuracy. As a result, it is essential that the algorithms' accuracy, recall, and F-measure are all high.

Keyphrases: TensorFlow, database, deep learning

@booklet{EasyChair:10523,
  author    = {Arju Kumar and Saurav Kumar and Kishan Kumar and Bharat Bhushan Naib},
  title     = {Email Fraud Detection},
  howpublished = {EasyChair Preprint 10523},
  year      = {EasyChair, 2023}}