Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fake Artificial Intelligence Generated Contents (FAIGC): A Survey of Theories, Detection Methods, and Opportunities

Xiaomin Yu, Yezhaohui Wang, Yanfang Chen, Zhen Tao, Dinghao Xi, Shichao Song, Simin Niu, Zhiyu Li

TL;DR

The paper addresses FAIGC by proposing a comprehensive taxonomy that separates intent (disinformation vs misinformation), modalities (text, visual, audio, multimodal), and creation methods (generation vs editing). It surveys AI-generated disinformation, AI-generated misinformation (hallucinations), and detection approaches across modalities, including text prompts, jailbreaks, diffusion-based generation, and Deepfake technologies. It reviews detection methods and benchmarks for deception, hallucination, and deepfake content, and discusses datasets, evaluation protocols, and practical challenges. The work highlights future directions such as multimodal FAIGC detection, zero-shot learning, interpretability, continual learning, and fact-verification-based detection, aiming to guide robust, trustworthy AI deployment in real-world settings.

Abstract

In recent years, generative artificial intelligence models, represented by Large Language Models (LLMs) and Diffusion Models (DMs), have revolutionized content production methods. These artificial intelligence-generated content (AIGC) have become deeply embedded in various aspects of daily life and work. However, these technologies have also led to the emergence of Fake Artificial Intelligence Generated Content (FAIGC), posing new challenges in distinguishing genuine information. It is crucial to recognize that AIGC technology is akin to a double-edged sword; its potent generative capabilities, while beneficial, also pose risks for the creation and dissemination of FAIGC. In this survey, We propose a new taxonomy that provides a more comprehensive breakdown of the space of FAIGC methods today. Next, we explore the modalities and generative technologies of FAIGC. We introduce FAIGC detection methods and summarize the related benchmark from various perspectives. Finally, we discuss outstanding challenges and promising areas for future research.

Fake Artificial Intelligence Generated Contents (FAIGC): A Survey of Theories, Detection Methods, and Opportunities

TL;DR

The paper addresses FAIGC by proposing a comprehensive taxonomy that separates intent (disinformation vs misinformation), modalities (text, visual, audio, multimodal), and creation methods (generation vs editing). It surveys AI-generated disinformation, AI-generated misinformation (hallucinations), and detection approaches across modalities, including text prompts, jailbreaks, diffusion-based generation, and Deepfake technologies. It reviews detection methods and benchmarks for deception, hallucination, and deepfake content, and discusses datasets, evaluation protocols, and practical challenges. The work highlights future directions such as multimodal FAIGC detection, zero-shot learning, interpretability, continual learning, and fact-verification-based detection, aiming to guide robust, trustworthy AI deployment in real-world settings.

Abstract

In recent years, generative artificial intelligence models, represented by Large Language Models (LLMs) and Diffusion Models (DMs), have revolutionized content production methods. These artificial intelligence-generated content (AIGC) have become deeply embedded in various aspects of daily life and work. However, these technologies have also led to the emergence of Fake Artificial Intelligence Generated Content (FAIGC), posing new challenges in distinguishing genuine information. It is crucial to recognize that AIGC technology is akin to a double-edged sword; its potent generative capabilities, while beneficial, also pose risks for the creation and dissemination of FAIGC. In this survey, We propose a new taxonomy that provides a more comprehensive breakdown of the space of FAIGC methods today. Next, we explore the modalities and generative technologies of FAIGC. We introduce FAIGC detection methods and summarize the related benchmark from various perspectives. Finally, we discuss outstanding challenges and promising areas for future research.
Paper Structure (69 sections, 10 figures, 2 tables)

This paper contains 69 sections, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. FAIGC Taxonomy
  3. The Intent Behind FAIGC
  4. The Modalities and Generative Technologies of FAIGC
  5. The Creation Method of FAIGC.
  6. AI-Generated Disinformation
  7. Text Modality
  8. LLMs Generation Approaches
  9. Jailbreak Technology
  10. Visual Modality
  11. Image
  12. 3.2.1.1 Controllable Image Generation.
  13. 3.2.1.2 Image Editing
  14. Video
  15. 3.2.2.1 Controllable Video Generation.
  16. ...and 54 more sections

Figures (10)

  • Figure 1: Typical AIGC examples: text generation, image generation, and deepfake in videos.
  • Figure 2: This line chart clearly illustrates that the search trends related to AI forgery, or AI crime show a strong positive correlation with the search trends for AIGC. This serves as a reminder that while focusing on the innovations brought about by AIGC, we must also further investigate the potential risks associated with FAIGC.
  • Figure 3: The overall framework of the FAIGC taxonomy. This framework focuses on three aspects: the intent behind FAIGC, the modalities and the generative technologies of FAIGC, and the creation method of FAIGC.
  • Figure 4: The structure of Sec.\ref{['sec3:AI-Generated Disinformation']} AI-generated Disinformation and Sec.\ref{['sec4:AI-Generated Misinformation']} AI-generated Misinformation.
  • Figure 5: A typical jailbreak behavior: constructing a hypothetical scenario to induce LLMs to generate harmful content.
  • ...and 5 more figures