Nordin, Nabil and Badrol, Muhamad Hasiff and Mansor, Hafizah (2025) Detectify: spam detection message app using NLP and ML. In: Selangor Techsphere Innovation Competition (STIC 2025), KLCC.
![[img]](https://irep.iium.edu.my/style/images/fileicons/application_pdf.png) |
PDF
- Published Version
Restricted to Repository staff only Download (782kB) | Request a copy |
Abstract
Spam messages have been surging through digital communication platforms such as WhatsApp, Telegram, and other mobile messaging apps that can possibly deceive users, particularly those less familiar with technology, into revealing their private information. Detectify is an innovative mobile application developed to address this issue by integrating intelligent detection mechanisms to identify potential scam messages. This app has two types of inputs for accessibility and user friendliness, which are text input and a screenshot of messages that can be scanned and transformed into text using Op cal Character Recognition (OCR) processing. This article focuses on the findings of a study looking at the accuracy results of different types of machine learning models and choosing the best machine learning model for implementation. The assessment involved a comparison between two types of learning models, which are the traditional machine learning model, including Naïve Bayes, Logis c Regression, Support Vector Machine (SVM), Random Forest, XGBoost, and the deep learning model, namely BERT and DistilBERT. The chosen traditional machine learning models are tested using TF-IDF vectorisation, and deep learning models using transformer embeddings. These tests are conducted on a compiled dataset of over 5000 messages, labelled as “scam” or “legit”. Experimental results demonstrated that DistilBERT outperformed other models, achieving an accuracy exceeding 90% and an F1-score above 0.89, indica ng superior contextual understanding and classification performance. Based on these findings, DistilBERT was selected for deployment in Detectify. The innovation on lies in combining OCR processing with transformer-based natural language processing (NLP) techniques to deliver an efficient and user-friendly spam detection solution. This integration enables Detectify to provide real- me scam message identification, thereby enhancing digital safety and fostering greater user confidence in mobile communication environments.
| Item Type: | Proceeding Paper (Poster) |
|---|---|
| Uncontrolled Keywords: | Spam, Scam message, Machine learning model, OCR, Deep learning model, Communication, NLP, Implementation |
| Subjects: | T Technology > T Technology (General) |
| Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button): | Kulliyyah of Information and Communication Technology > Department of Computer Science Kulliyyah of Information and Communication Technology > Department of Computer Science |
| Depositing User: | Hafizah Mansor |
| Date Deposited: | 10 Aug 2026 23:59 |
| Last Update: | 10 Aug 2026 23:59 |
| Queue Number: | 2026-08-Q4691 |
| URI: | http://irep.iium.edu.my/id/eprint/130383 |
Actions (login required)
 |
View Item |