Deepfake Detector: Explainable Detection of AI-Generated Face Manipulations Using CNN and Grad-CAM

Zeeshan Ali; Muhammad Mudassir; Amanat Ali; Muhammad Hammad Khan

Authors

Zeeshan Ali Department of Computer Science Quaid-e-Awam University of Engineering, Science and Technology, Nawabshah, Sindh, Pakistan
Muhammad Mudassir Department of Computer Science Quaid-e-Awam University of Engineering, Science and Technology, Nawabshah, Sindh, Pakistan
Amanat Ali Department of Computer Science Quaid-e-Awam University of Engineering, Science and Technology, Nawabshah, Sindh, Pakistan
Muhammad Hammad Khan Department of Computer Science Quaid-e-Awam University of Engineering, Science and Technology, Nawabshah, Sindh, Pakistan

Keywords:

Deepfake Detection, Convolutional Neural Networks, Explainable Artificial Intelligence, Grad-CAM, Image Forgery Detection

Abstract

Today’s AI technology enables the creation of fake images that look incredibly real. These AI-generated images appear very realistic, making it difficult to distinguish between original photos and AI-generated images. Unfortunately, such AI technology can also be misused for identity theft, spreading misinformation, or harming someone’s reputation. Therefore, finding effective ways to recognize these forged photos becomes essential. In this research, we introduce a deep learning system that uses a Convolutional Neural Network (CNN) to analyze images and identify subtle tampering artifacts to the human eye, thereby referred to as a deepfake detection system. Unlike existing detection systems, which only determine whether an image is "fake" or "real", this system features Grad-CAM, an explainable AI technique that produces color-coded heat maps to identify the precise parts of a photo, such as eyes, mouth, or skin, that influenced the model’s decision. Our approach also ensures honesty and interpretability, enabling people to accept the result through the integration of explainable graphics and deep learning. We want to create a user-friendly resource to help people verify images and reduce the spread of misinformation online.

References

Nazneen Mansoor, Alexander I. Iliev, “Explainable AI for DeepFake Detection,” Agri Sci., vol. 15, no. 2, p. 725, 2025, doi: https://doi.org/10.3390/app15020725.

Amritha N. Devi, Philomina Simon, “DeepGuardNet: : A Novel CNN Architecture for DeepFake Image Detection,” Procedia Comput. Sci., vol. 258, 2025, doi: https://doi.org/10.1016/j.procs.2025.04.313.

Mohd Tahir Irfan, Bhavna Arora, “On Machine Learning and Deep Learning based Deepfake Generation and Detection,” Procedia Comput. Sci., vol. 259, pp. 1927–1936, 2025, doi: https://doi.org/10.1016/j.procs.2025.04.148.

Sonya J. Burroughs, Balakrishna Gokaraju, “Exposing deepfakes using a deep multilayer perceptron – convolutional neural network model,” Int. J. Inf. Manag. Data Insights, vol. 2, no. 1, p. 100054, 2022, doi: https://doi.org/10.1016/j.jjimei.2021.100054.

Muhammad Asad Arshed, Shahzad Mumtaz, “Multiclass AI-Generated Deepfake Face Detection Using Patch-Wise Deep Learning Model,” Computers, vol. 13, no. 1, p. 31, 2024, doi: https://doi.org/10.3390/computers13010031.

K. Kapoor, S. P. Singh, G. Aggarwal, and M. Sharma, “Deepfake Detection Using CNN-Based Architecture,” Int. Conf. Eng. Technol. Manag. ICETM 2025, 2025, doi: 10.1109/ICETM63734.2025.11051404.

Nurcan Yardımcı, Mohamed Ibrahim Abdi, “Hibrit ResNeXt ve LSTM Mimarisi Kullanılarak Deepfake Video Algılama,” Politek. Derg., 2025, doi: 10.2339/politeknik.1721371.