DEEPFAKE DETECTION OF AI IMAGES AND AUDIO USING DEEP LEARNING

This review has examined deep learning approaches for detecting potentially AI-generated images and audio and outlined a framework that brings both tasks into one application. Image CNNs can learn spatial features such as texture, boundaries and visual inconsistencies, whereas audio models can analyse waveform or spectrogram features and use temporal networks to represent changes across speech. By applying separate preprocessing steps and modality-specific models, the proposed system can accept either type of media and present a real/fake prediction with a confidence estimate through a common result interface. This design makes the detection workflow easier to use while respecting the different characteristics of image and audio data. The reliability of such a system depends not only on its model architecture but also on the quality and diversity of its datasets, the separation of training and testing data, and the realism of evaluation conditions. Compression, noise, unfamiliar generators and dataset bias can reduce performance, so evaluation should use independent samples and several metrics rather than a single accuracy figure. Confidence scores should be communicated carefully, and an inconclusive result should remain possible when the evidence is weak. Since synthetic-media techniques continue to evolve, ongoing testing and updating are necessary. Overall, deep learning can provide useful support for preliminary deepfake screening, but it should not replace human judgement, provenance checks or expert forensic analysis. Responsible data handling and transparent limitations are essential for practical use.

[1]   H.-H. Nguyen-Le et al., “Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness,” Artificial Intelligence Review, vol. 59, article 203, 2026. [2]   “A Survey on Speech Deepfake Detection,” ACM Computing Surveys, vol. 57, no. 7, article 165, 2025, doi: 10.1145/3714458. [3]   J. Yi et al., “Audio Deepfake Detection: A Survey,” arXiv:2308.14970, 2023. [4]   A. Rössler et al., “FaceForensics++: Learning to Detect Manipulated Facial Images,” Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019. [5]   B. Dolhansky et al., “The DeepFake Detection Challenge (DFDC) Dataset,” arXiv:2006.07397, 2020. [6]   ASVspoof, “Automatic Speaker Verification Spoofing and Countermeasures Challenge,” challenge datasets and evaluation resources. [7]   “Deepfake: definitions, performance metrics and standards, datasets, and a meta-review,” available through PubMed Central, 2024.