The emergence of deep fake technologies has prompted serious concerns regarding the authenticity and trustworthiness of digital media. This paper addresses the challenge of deep fake image detection using a combination of generative adversarial networks (GANs) for synthetic image generation and convolutional neural networks (CNNs) for classification. A custom dataset comprising 5000 real-face images and synthetic images generated using GANs is used for training and evaluation. Our approach leverages transfer learning by fine-tuning a DenseNet121 architecture pre-trained on ImageNet for improved feature extraction and classification performance. The model is trained and validated using a split of 70% training, 10% validation, and 20% testing datasets. The training process is evaluated through metrics including accuracy, loss, and receiver operating characteristic (ROC) curve analysis. The proposed methodology demonstrates promising results in discerning between real and fake images, achieving high accuracy in distinguishing manipulated images from genuine ones. This research contributes to the ongoing efforts in combating the adverse effects of deep fake technologies on various domains, including media, politics, and cybersecurity.