Detecting Ai-generated Content: A Survey On Multimodal Detection Of Text, Image, And Video

• Author(s):

  Davis Jacob K | Mukesh M Suthar | Sohara Banu AR

• Keywords:

  Generative AI, Multimodal Detection, AI-Generated Content, Cross-Modal Consistency, Adversarial Training, Text Detection, Image Detection, Video Detection, Explainable AI, Dataset Diversity, Scalability.

• Abstract:

  Generative AI Has Advanced Very Quickly Allowing The Generation Of Realistic Fake Content In Text, Image, And Video Domains That Is Becoming Challenging To Differentiate From Real Content. While There Has Been Some Progress In Developing Detectors That Work Within Specific Modalities To Identify The AI-generated Content, All Of Them Suffer From The Lack Of Exploitation Of Inter-modal Contradiction And Dependencies. In This Survey, The Drawbacks Of Existing Systems Are Described From The Points Of View Of Scalability, Robustness, The Variety Of Datasets Used, And Overall Efficiency. It Then Formulates A New Scheme For A Multimodal Detection Mechanism That Can Detect Text, Image As Well As Videos All At Once. To Improve The Detection Accuracy, Scalability And Use Across Broad Cultural And Language Settings This Framework Utilizes —Efficient Multimodal Models, Cross Modal Consistency Checks, Adversarial Training, And Efficient Architecture. The Proposed System Offsets The Gap Between Standard Approaches And The Innovative Advancement Of Multimodal Generative AI By Providing A Real-time Detection System For Adversarial Signals. It Provides A Unifying Model That Underpins Solid, Context-sensitive Detection Schemes To Protect Society’s Trust While Preventing The Abuse Of Generative AI.

---

Other Details

• Paper id:

  IJSARTV11I4103172

• Published in:

  Volume: 11 Issue: 4 April 2025

• Publication Date:

  2025-04-16

Download Article