Allen Varghese


Presentation: 11:40 a.m.-12:00 p.m., Jessie Hathcock Hall 101



Download Project (1.5 MB)


The increasing use of web-based applications has led to agrowing need for robust and secure systems that can ensure the privacy and security of sensitive information. Unfortunately, the functions and APIs used by these applications are often complex and prone to exploitation, making it difficult to detect and prevent malicious activity. To address these challenges, we propose a deep learning-based approach that detects malicious behaviors at run time.The proposed approach leverages APIs and function call at runtime to detect malicious behaviors. More specifically, we trained a deep learning model on the data extracted from 1 million web apps. The use of deep learning to monitor these functions is a novel approach that has the potential to provide real-time protection against malicious activities.Implementing the proposed solution involves writing a JavaScript script that modifies the monitored functions. The script assigns each function to a new custom function that logs its usage and calls the original function. The custom functions use the apply method to preserve the context of the original function. The information collected from logging the functions is then used to train the machine learning model. The expected outcome of this thesis is to deliver a functional implementation of the proposed framework that can effectively detect malicious activities, while also generating useful usage insights for JavaScript APIs.

Publication Date


Project Designation

Graduate Research

Primary Advisor

Phu Phung

Primary Advisor's Department

Computer Science


Stander Symposium, College of Arts and Sciences

Institutional Learning Goals

Scholarship; Practical Wisdom; Critical Evaluation of Our Times

Dynamic Analysis Framework for Classifying Malicious Web Pages