Privacy & Transparency

We use cookies to secure the credit system and serve personalized ads (Google AdSense). Your uploaded media is never stored.

Forensic Science

Detection Methodology

AI-generated images are statistically distinguishable from real photographs - not visually, but mathematically. FauxLens applies six independent forensic signals drawn from academic research and production-tested on real-world synthetic media, achieving 98.1% detection accuracy across 1.2M+ analyzed images. Here is exactly how each technique works.

Analysis Pipeline

When a camera captures a photograph and saves it as JPEG, the entire image is compressed simultaneously using a Discrete Cosine Transform (DCT) algorithm operating on 8×8 pixel blocks. Every region of that image shares the same compression generation - the same "quality history." AI-generated images and composites do not share this uniform history. When a diffusion model renders an image, it generates pixels through iterative denoising rather than optical capture. When an editor pastes an AI face onto a real photograph, the spliced region carries different compression artifacts than the original. ELA works by re-saving the image at a known quality level (typically 95%) and computing the absolute difference between the re-saved version and the original. In a pristine photograph, this difference map - the ELA map - is relatively uniform across all regions. In a manipulated or AI-generated image, regions with different compression histories light up with elevated error values, appearing bright in the ELA visualization against a darker background. This technique is particularly effective for detecting image compositing. A face with a dramatically different ELA response than the surrounding background is a strong signal of digital insertion. Pure AI-generated images also exhibit ELA patterns characteristic of the generator's output format - typically lacking the graduated compression artifacts that distinguish real photographic content. Important limitation: images that have been re-compressed multiple times (for example, saved through social media platforms which re-encode uploaded files) accumulate compression artifacts that can complicate ELA interpretation. This is why ELA is used as one signal among several rather than as a standalone verdict.

Accuracy, Limitations, and Responsible Use

No forensic technique is 100% accurate. Our multi-signal pipeline achieves high accuracy on high-quality AI-generated images but has known limitations: heavily re-compressed images degrade ELA evidence, images with no visible shadows limit shadow logic analysis, and adversarially trained models may suppress GAN fingerprints.

Results should be treated as probabilistic evidence - a strong signal warranting further investigation - not as definitive verdicts. We present confidence scores and contributing evidence chains specifically so that you, the human evaluator, can apply context and judgment that no algorithm possesses. Detection results should never be the sole basis for legal action, editorial decisions, or reputational judgments.

All images submitted are processed ephemerally and deleted immediately after analysis. We do not retain, index, or train on user-submitted images. Read more about our ethics and data policy.

Academic References

Peer-reviewed research underpinning each detection technique

  1. [1]Krawetz, N. (2007). "A Picture's Worth: Digital Image Analysis and Forensics." Black Hat Briefings. - Foundational paper on JPEG Error Level Analysis (ELA).
  2. [2]Lukas, J., Fridrich, J., & Goljan, M. (2006). "Digital Camera Identification from Sensor Pattern Noise." IEEE Transactions on Information Forensics and Security, 1(2), 205-214. - PRNU noise fingerprinting methodology.
  3. [3]Frank, J., Eisenhofer, T., Schönherr, L., Fischer, A., Kolossa, D., & Holz, T. (2020). "Leveraging Frequency Analysis for Deep Fake Image Recognition." Proceedings of ICML 2020. arXiv:2003.08685. - Fourier frequency domain artifacts in GAN/diffusion outputs.
  4. [4]Wang, S., Wang, O., Zhang, R., Owens, A., & Efros, A. A. (2020). "CNN-Generated Images Are Surprisingly Easy to Spot... For Now." CVPR 2020. arXiv:1912.11035. - GAN fingerprint generalization across architectures.
  5. [5]Rössler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., & Nießner, M. (2019). "FaceForensics++: Learning to Detect Manipulated Facial Images." ICCV 2019. arXiv:1901.08971. - Deepfake detection benchmark and dataset.
  6. [6]C2PA (Coalition for Content Provenance and Authenticity). (2023). "C2PA Technical Specification v1.3." c2pa.org - Content credentials and provenance metadata standard.
  7. [7]Marra, F., Gragnaniello, D., Cozzolino, D., & Verdoliva, L. (2019). "Do GANs Leave Artificial Fingerprints?" IEEE MIPR 2019. arXiv:1812.11842. - Architecture-specific GAN fingerprint extraction.

Ready to put the methodology into practice?

Run a Free Analysis