Detecting Fraudulent/Scam Phone Calls using Machine Learning

Local Chapter Singapore Alpha Chapter

Coordinated bySingapore ,

Choon Xiang Chwa
, Pow Look Yap

Status: Completed

Project Duration: 09 Apr 2023 - 11 Jun 2023

Open Source resources available from this project

Github Repository

Project background.

Cases of scam or fraudulent phone calls have been escalating all around the world. In the US, the consumers fend off more than 3b scam calls a month with Americans losing $30b to scam calls in the first half of 2021 alone. In Singapore itself, the reported number of scam calls in first half of 2022 spiked to 13,000 cases, an increase of almost 100% over the previous year. The total sum lost to scam calls amount to $347m. According to a consumer report, almost 90% of Singapore consumers have received scam calls. What is interesting in the finding is that 75% continue to pick up the calls, despite the extensive coverage of this issue in the mainstream media.

The problem.

Spam calls are very difficult to prevent despite the efforts of the authorities and telcos. The scammers always manage to find ways to circumvent for their next round for every solution that the authorities apply. It can be said that it is a proverbial dog chasing the tail problem. Nonetheless, there will continue to be efforts to try to solve this problem.
 
In this project,  we aim to find ways to detect fraudulent phone calls before they pick up or while they are still taking and alerting the user. We plan to perform the detection using voice prints, robotic voice identification as well as using machine learning on the phone content. We will develop an app that will give access to the call to perform recognition and therefore alert the users.

Project goals.

Our goal is to develop an app that will detect fraudulent calls using machine learning. This app can be installed to a mobile phone whereby it will be able to inform the user if the call they receive is fraudulent or not.  With a duration of 10-weeks, this project aims to do the following:   1. Data Collection and Exploratory Data Analysis - Collect voice-data on scam calls - Collect voice-data on robotic voice vs human voice 2. Pre-processing 3. Feature Extraction   4. Model Development and Training 5. Evaluate Model 6. App development

Project plan.

  • Week 1

    Research previous work and Data Collection

  • Week 3

    Data Collection

  • Week 4

    Exploratory Data Analysis

  • Week 5

    Pre-processing and Augmentation

  • Week 6

    Model Development

  • Week 7

    Model Training

  • Week 8

    Model Analysis and Interpretation. Week 9-10: App Development

Learning outcomes.

The participants will be exposed to the complete pipeline of machine learning development from conception to app development

Share project on: