Detect-LAIM-generated-Multimedia-Survey

This repository contains a collection of resources and papers on Detecting Multimedia Generated by Large AI Models: A Survey

The references of those works are displayed in Generation Works and Detection Works.

Please let us know if you find a mistake, or if we have missed your wonderful work by e-mail: lin1785@purdue.edu, hu968@purdue.edu, gupt1031@purdue.edu

If you find our survey useful for your research, please cite the following Paper

@article{lin2024detecting,
  title={Detecting Multimedia Generated by Large AI Models: A Survey},
  author={Lin, Li and Gupta, Neeraj and Zhang, Yue and Ren, Hainan and Liu, Chun-Hao and Ding, Feng and Wang, Xin and Li, Xin and Verdoliva, Luisa and Hu, Shu},
  journal={arXiv preprint arXiv:2402.00045},
  year={2024}
}

💻 Contents

• Related Works
• Generation
  • Public Datasets for Detection
• Detection
  • Text
    • Pure Detection
      • Easy Explainable Methods
        • Watermarking
        • Non-Watermarking
      • Hard Explainable Methods
    • Beyond Detection
      • Efficiency
      • Attribution
      • Generalization
      • Interpretability
      • Robustness
        • Adversarial Attack Robustness
        • LAIM-Polished Robustness
      • Empirical Study
  • Image
    • Pure Detection
      • Physical/Physiological based Methods
      • Diffuser Fingerprints based Methods
      • Spatial-based Methods
      • Frequency-based Methods
      • Distribution-based Methods
    • Beyond Detection
      • Attribution and Model Parsing
      • Generalization
      • Interpretability
      • Localization
        • Fully-supervised
        • Weakly-supervised
      • Robustness
        • Adversarial Attack Robustness
        • Post-Processing Robustness
      • Empirical Study
  • Video
    • Pure Detection
      • Spatial&Temporal-based Methods
    • Beyond Detection
      • Generalization
      • Empirical Study
  • Audio
    • Pure Detection
      • Vocoder-based
    • Beyond Detection
      • Generalization
  • Multimodal
    • Pure Detection
      • Prompt-guided
      • Text-image Inconsistency
    • Beyond Detection
      • Attribution
      • Generalization
        • Prompt Tuning
        • Contrastive Learning
      • Interpretability
      • Localization
        • Spatial-based
        • Frequency-based
        • MLLM-based
      • Empirical Study
• Detection Tools

📈 Related Work

         - A Survey on Detection of LLMs-Generated Content Paper GitHub

         - A Survey on LLM-generated Text Detection: Necessity, Methods, and Future Directions Paper GitHub

         - Towards possibilities & impossibilities of ai-generated text detection: A survey Paper

         - Machine-generated text: A comprehensive survey of threat models and detection methods Paper

         - The Age of Synthetic Realities: Challenges and Opportunities Paper

         - GenAI against humanity: Nefarious applications of generative artificial intelligence and large language models Paper

         - A Comprehensive Survey of Fake Text Detection on Misinformation and LM-Generated Texts Paper

         - Recent Advances on Generalizable Diffusion-generated Image Detection Paper

         - Survey on AI-Generated Media Detection: From Non-MLLM to MLLM Paper

         - Passive Deepfake Detection Across Multi-modalities: A Comprehensive Survey Paper

Generation

Illustrations of different types of multimedia generation process based on LAIMs.

Public Datasets for Detection

Please read the column I20(Input-to-Output) with these abbreviations:

• T2T: Text-to-Text
• V2T: Video-to-Text
• T2I: Text-to-Image
• I2I: Image-to-Image
• T2A: Text-to-Audio
• I.A2V: (Image conditioned with Audio)-to-Video

:mag_right: Detection :fire:

:page_facing_up: Text

---

Pure Detection

Illustrations of pure detection methodologies for LAIM-generated text.

  ♣️ Easy Explainable Methods

        ▶️ Watermarking

         - Distillation-Resistant Watermarking for Model Protection in NLP Paper

         - Three bricks to consolidate watermarks for large language models Paper GitHub

         - Robust multi-bit natural language watermarking through invariant features Paper

         - Undetectable Watermarks for Language Models Paper

         - Robust distortion-free watermarks for language models Paper

         - Provable robust watermarking for ai-generated text Paper GitHub

         - A Private Watermark for Large Language Models Paper

        ▶️ Non-watermarking

         - Unraveling the mystery of artifacts in machine generated text Paper

         - Stylometric detection of ai-generated text in twitter timelines Paper

         - CoCo: Coherence-Enhanced Machine-Generated Text Detection Under Data Limitation With Contrastive Learning Paper

         - Beat LLMs at Their Own Game: Zero-Shot LLM-Generated Text Detection via Querying ChatGPT Paper GitHub

         - Who Wrote This? The Key to Zero-Shot LLM-Generated Text Detection Is GECScore Paper GitHub

  ♣️ Hard Explainable Methods

         - HowkGPT: Investigating the Detection of ChatGPT-generated University Student Homework through Context-Aware Perplexity Analysis Paper

         - GPTZero Tool

         - Detectgpt: Zero-shot machine-generated text detection using probability curvature Paper GitHub

         - Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text Paper GitHub

         - Multiscale Positive-Unlabeled Detection of AI-Generated Texts Paper GitHub

Beyond Detection

Illustrations of beyond detection methodologies for LAIM-generated text.

  ♣️ Efficiency

         - Efficient Detection of LLM-generated Texts with a Bayesian Surrogate Model Paper

         - Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature Paper GitHub

         - DetectLLM: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text Paper GitHub

         - SeqXGPT: Sentence-Level AI-Generated Text Detection Paper GitHub

         - Glimpse: Enabling White-Box Methods to Use Proprietary Models for Zero-Shot LLM-Generated Text Detection Paper GitHub

  ♣️ Attribution

         - TURINGBENCH: A Benchmark Environment for Turing Test in the Age of Neural Text Generation Paper Turingbench

         - Whodunit? Learning to Contrast for Authorship Attribution Paper

         - Through the looking glass: Learning to attribute synthetic text generated by language models Paper

         - TopRoBERTa: Topology-Aware Authorship Attribution of Deepfake Texts Paper

         - Authorship attribution for neural text generation Paper GitHub

         - Gpt-who: An information density-based machine-generated text detector Paper

         - LLMDet: A Third Party Large Language Models Generated Text Detection Tool Paper GitHub

         - Few-Shot Detection of Machine-Generated Text using Style Representations Paper

         - Origin Tracing and Detecting of LLMs Paper

  ♣️ Generalization

         - Ghostbuster: Detecting Text Ghostwritten by Large Language Models Paper

         - Conda: Contrastive domain adaptation for ai-generated text detection Paper GitHub

         - Text Fluoroscopy: Detecting LLM-Generated Text through Intrinsic Features Paper GitHub

         - DeTeCtive: Detecting AI-generated Text via Multi-Level Contrastive Learning Paper GitHub

         - Intrinsic Dimension Estimation for Robust Detection of AI-Generated Texts Paper GitHub

  ♣️ Interpretability

         - DNA-GPT: Divergent N-Gram Analysis for Training-Free Detection of GPT-Generated Text Paper GitHub

         - A Watermark for Large Language Models Paper GitHub

         - Chatgpt or human? detect and explain. explaining decisions of machine learning model for detecting short chatgpt-generated text Paper

         - Check Me If You Can: Detecting ChatGPT-Generated Academic Writing using CheckGPT Paper

         - Is chatgpt involved in texts? measure the polish ratio to detect chatgpt-generated text Paper

  ♣️ Robustness

        ▶️ Adversarial Attack Robustness

         - Red Teaming Language Model Detectors with Language Models Paper

         - Radar: Robust ai-text detection via adversarial learning Paper Project Page

         - J-guard: Journalism guided adversarially robust detection of ai-generated news Paper

         - Outfox: Llm-generated essay detection through in-context learning with adversarially generated examples Paper

        ▶️ LAIM-Polished Robustness

         - Is chatgpt involved in texts? measure the polish ratio to detect chatgpt-generated text Paper

  ♣️ Empirical Study

         - ChatLog: Recording and Analyzing ChatGPT Across Time Paper GitHub

         - On the Zero-Shot Generalization of Machine-Generated Text Detectors Paper

         - On the Generalization of Training-based ChatGPT Detection Methods Paper

         - Supervised Machine-Generated Text Detectors: Family and Scale Matters Paper GitHub

         - Deepfake Text Detection in the Wild Paper GitHub

         - How large language models are transforming machine-paraphrased plagiarism Paper

         - Paraphrase Detection: Human vs. Machine Content Paper

         - MGTBench: Benchmarking Machine-Generated Text Detection Paper GitHub

         - How close is chatgpt to human experts? comparison corpus, evaluation, and detection Paper GitHub

         - Can LLM-Generated Misinformation Be Detected? Paper GitHub

         - From Text to Source: Results in Detecting Large Language Model-Generated Content Paper

📸 Image

---

Pure Detection

Illustrations of pure detection methodologies for LAIM-generated image.

  ♣️ Physical/Physiological based Methods

         - Qualitative Failures of Image Generation Models and Their Application in Detecting Deepfakes Paper

         - Perspective (in) consistency of paint by text Paper

         - Lighting (in) consistency of paint by text Paper

  ♣️ Diffuser Fingerprints based Methods

         - Deep Image Fingerprint: Accurate And Low Budget Synthetic Image Detector Paper

         - DIRE for Diffusion-Generated Image Detection Paper GitHub

         - Exposing the Fake: Effective Diffusion-Generated Images Detection Paper

         - LaRE^2: Latent Reconstruction Error Based Method for Diffusion-Generated Image Detection Paper GitHub

         - Aligned Datasets Improve Detection of Latent Diffusion-Generated Images Paper GitHub

         - Manifold Induced Biases for Zero-shot and Few-shot Detection of Generated Images Paper GitHub

  ♣️ Spatial-based Methods

         - Rich and Poor Texture Contrast: A Simple yet Effective Approach for AI-generated Image Detection Paper Project Page

         - Unmasking The Artist: Discriminating Human-Drawn And AI-Generated Human Face Art Through Facial Feature Analysis Paper

         - Detecting images generated by deep diffusion models using their local intrinsic dimensionality Paper

  ♣️ Frequency-based Methods

         - Wavelet-packets for deepfake image analysis and detection Paper GitHub

         - AUSOME: authenticating social media images using frequency analysis Paper

         - AI-Generated Image Detection using a Cross-Attention Enhanced Dual-Stream Network Paper

         - Synthbuster: Towards Detection of Diffusion Model Generated Images Paper

         - Faster Than Lies: Real-time Deepfake Detection using Binary Neural Networks Paper GitHub

  ♣️ Distribution-based Methods

         - Zero-Shot Detection of AI-Generated Images Paper GitHub

Beyond Detection

Illustrations of beyond detection methodologies for LAIM-generated image.

  ♣️ Attribution and Model Parsing

        ▶️ Attribution and Model Parsing

         - Level up the deepfake detection: a method to effectively discriminate images generated by gan architectures and diffusion models Paper

         - Reverse engineering of generative models: Inferring model hyperparameters from generated images Paper

  ♣️ Generalization

         - Online Detection of AI-Generated Images Paper

         - Towards universal fake image detectors that generalize across generative models Paper GitHub

         - Raising the Bar of AI-generated Image Detection with CLIP Paper

         - Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection Paper

         - Fingerprintnet: Synthesized fingerprints for generated image detection Paper

         - Detecting Deepfakes Without Seeing Any Paper GitHub

         - Improving Synthetically Generated Image Detection in Cross-Concept Settings Paper

         - Diffusion Noise Feature: Accurate and Fast Generated Image Detection Paper

         - Contrasting Deepfakes Diffusion via Contrastive Learning and Global-Local Similarities Paper GitHub

         - HRR: Hierarchical Retrospection Refinement for Generated Image Detection Paper

         - A Sanity Check for AI-generated Image Detection Paper GitHub

         - Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face Forgery Detection Paper GitHub

         - A Bias-Free Training Paradigm for More General AI-generated Image Detection Paper GitHub

         - Breaking Semantic Artifacts for Generalized AI-generated Image Detection Paper GitHub

         - Dual Data Alignment Makes AI‑Generated Image Detector Easier Generalizable Paper

  ♣️ Interpretability

         - Interpretable-through-prototypes deepfake detection for diffusion models Paper GitHub

         - Did You Note My Palette? Unveiling Synthetic Images Through Color Statistics Paper

  ♣️ Localization

        ▶️ Fully-supervised

         - Hierarchical fine-grained image forgery detection and localization Paper GitHub

         - Perceptual Artifacts Localization for Image Synthesis Tasks Paper GitHub

         - TruFor: Leveraging all-round clues for trustworthy image forgery detection and localization Paper GitHub

         - UnionFormer: Unified-Learning Transformer with Multi-View Representation for Image Manipulation Detection and Localization Paper

        ▶️ Weakly-supervised

         - Weakly-supervised deepfake localization in diffusion-generated images Paper

  ♣️ Robustness

        ▶️ Adversarial Attack Robustness

         - D4: Detection of Adversarial Diffusion Deepfakes Using Disjoint Ensembles Paper

         - Exploring the Adversarial Robustness of CLIP for AI-generated Image Detection Paper

         - All Patches Matter, More Patches Better: Enhance AI‑Generated Image Detection via Panoptic Patch Learning Paper

        ▶️ Post-Processing Robustness

         - GLFF: Global and Local Feature Fusion for AI-synthesized Image Detection Paper

         - Exposing fake images generated by text-to-image diffusion models Paper

         - Local Statistics for Generative Image Detection Paper

  ♣️ Empirical Study

         - On the detection of synthetic images generated by diffusion models Paper GitHub

         - Intriguing properties of synthetic images: from generative adversarial networks to diffusion models Paper

         - Towards the detection of diffusion model deepfakes Paper

         - Unveiling the Impact of Image Transformations on Deepfake Detection: An Experimental Analysis Paper

         - On the use of Stable Diffusion for creating realistic faces: from generation to detection Paper

         - Finding AI-Generated Faces in the Wild Paper

         - Forensic analysis of synthetically generated western blot images Paper

         - Beyond Human Forgeries: An Investigation into Detecting Diffusion-Generated Handwriting Paper

🎞️ Video

---

Illustration of detection methodology in generalization task for LAIM-generated video.

Pure Detection

  ♣️ Spatial & Temporal based Methods

         - Distinguish Any Fake Videos: Unleashing the Power of Large-scale Data and Motion Features Paper

         - Exposing AI-generated Videos: A Benchmark Dataset and a Local-and-Global Temporal Defect Based Detection Method Paper

Beyond Detection

  ♣️ Generalization

         - Revisiting Generalizability in Deepfake Detection: Improving Metrics and Stabilizing Transfer Paper

  ♣️ Empirical Study

         - Beyond Deepfake Images: Detecting AI-Generated Videos Paper

🎵 Audio

---

Pure Detection

The artifacts introduced by DM-based neural vocoders (WaveGrad and DiffWave) to a voice signal. The differences in mel-spectrograms between real and generated ones are illustrated in the third and fifth columns.

  ♣️ Vocoder-based

         - AI-Synthesized Voice Detection Using Neural Vocoder Artifacts Paper GitHub

Beyond Detection

  ♣️ Generalization

         - Improving Generalization for AI-Synthesized Voice Detection Paper GitHub

🍯 Multimodal

---

Pure Detection

Illustrations of pure detection methodologies for LAIM-generated multimodal media.

  ♣️ Prompt-guided

         - Parents and Children: Distinguishing Multimodal DeepFakes from Natural Images Paper

         - On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection Paper GitHub

         - Human Action CLIPS: Detecting AI-generated Human Motion Paper

  ♣️ Text-image Inconsistency

         - Detecting Cross-Modal Inconsistency to Defend Against Neural Fake News Paper GitHub

         - Exposing Text-Image Inconsistency Using Diffusion Models Paper

Beyond Detection

Illustrations of beyond detection methodologies for LAIM-generated multimodal media.

  ♣️ Attribution

         - De-fake: Detection and attribution of fake images generated by text-to-image generation models Paper

         - FIDAVL: Fake Image Detection and Attribution using Vision-Language Model Paper GitHub

  ♣️ Generalization

        ▶️ Prompt Tuning

         - AntifakePrompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors Paper GitHub

        ▶️ Contrastive Learning

         - Generalizable Synthetic Image Detection via Language-guided Contrastive Learning Paper GitHub

  ♣️ Interpretability

         - Combating Misinformation in the Era of Generative AI Models Paper

         - FFAA: Multimodal Large Language Model based Explainable Open-World Face Forgery Analysis Assistant Paper GitHub

         - X^2-DFD: A framework for eXplainable and eXtendable deepfake detection Paper

  ♣️ Localization

        ▶️ Spatial-based

         - Detecting and grounding multi-modal media manipulation Paper

         - Exploiting Modality-Specific Features For Multi-Modal Manipulation Detection And Grounding Paper

        ▶️ Frequency-based

         - Unified Frequency-Assisted Transformer Framework for Detecting and Grounding Multi-Modal Manipulation Paper

        ▶️ MLLM-based

         - FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models Paper GitHub

         - ForgeryGPT: Multimodal Large Language Model For Explainable Image Forgery Detection and Localization Paper

  ♣️ Empirical Study

         - Detecting Images Generated by Diffusers Paper GitHub

         - CLIPping the Deception: Adapting Vision-Language Models for Universal Deepfake Detection Paper

         - VERITE: a Robust benchmark for multimodal misinformation detection accounting for unimodal bias Paper GitHub

         - Can ChatGPT Detect DeepFakes? A Study of Using Multimodal Large Language Models for Media Forensics Paper

Detection Tools

Generation Works

---

Detection Works