Page 165 - tech fest 2025 ב״ש חוברת תקצירים

Deepfake in Photos
SE-A-13
Arbel Zagag; arbelar@ac.sce.ac.il Daniel Bar; danieba6@ac.sce.ac.il Yovel Nirenberg; yovelni@ac.sce.ac.il
Advisors: Dr. Irina Rabaev1, Ms. Alona Kutsyy1 1SCE - Shamoon College of Engineering, Be’er-Sheva
The rise of deepfake technology, enabled by generative adversarial networks (GANs), threatens digital authenticity by creating highly realistic fake images that are often indistinguishable to the human eye. This project focuses on developing a robust deepfake detection model using AI-based approaches, including frequency analysis, convolutional trace detection, and deep learning models. The research examines methods like convolutional neural networks (CNNs), support vector machines (SVMs), and hybrid detection models integrated with pre-trained architectures like VGG16, ResNet50, and Xception. The experimental results demonstrate that combining these methods improves performance, achieving high classification accuracy on datasets like OpenForensicsV1. With deepfake images used to spread misinformation, particularly in sensitive geopolitical contexts, this project contributes to developing a reliable, scalable system to combat digital manipulation.
Keywords: deepfake, deep-learning, GAN, images, models, photos
NetSpect - Detecting Cyberattacks in Network Traffic in Real Time
SE-A-14
Shay Hahiashvili; shayha2@ac.sce.ac.il Maxim Subotin; maximsu@ac.sce.ac.il
Advisor: Ms. Alona Kutsyy
SCE - Shamoon College of Engineering, Be'er-Sheva
In the modern era, dependence on technology and the internet is growing, and cyberattacks in the field of network communications pose serious risks to businesses and users, which can lead to data theft, service disruptions, and financial losses. As part of the project, we developed a real-time intrusion detection system (IDS) that monitors network traffic, detects attack patterns, and issues alerts. The hybrid system integrates cyberattack detection algorithms with machine learning models to identify anomalies. It detects threats like Port Scanning, DoS, ARP Spoofing, and DNS Tunneling. Since existing datasets were ineffective for real-time detection, we manually collected optimal training data. Our system ensures high accuracy with minimal false alarms and features an intuitive and simple interface.
Keywords: ARP Spoofing, denial of service, DNS Tunneling, intrusion detection system, machine learning, Port Scanning
Book of Abstracts | 2025
165