AI-Powered Forensics Platform
Is this image authentic?
ForgeGuard combines deep learning ensembles with classical forensics to detect AI-generated images, copy-move tampering, splicing, and inpainting — in seconds.
99%
DL Accuracy
4
Forgery Categories
11
CV Signal Types
Drop image here or browse
Drag & drop your image for instant analysis
↑ Choose File
JPEG
PNG
WEBP
Max 20MB
—
Analyzing Image
Loading image data
Running detection models
Analyzing signal patterns
Computing confidence scores
Generating forensic report
—
—
—
—
AI Image Detection
SigLIP2 ensemble trained on Midjourney, DALL-E 3, Flux.1, SD 3.5 and GPT-4o images.
Copy-Move Forensics
SIFT keypoint matching with RANSAC to detect cloned regions within an image.
Splicing Detection
JPEG ghost analysis and noise inconsistency mapping to find composite images.
Inpainting Detection
Texture repetition and over-smooth region analysis to locate AI-filled areas.
Process
How ForgeGuard Works
A two-engine hybrid approach combining state-of-the-art deep learning with classical digital forensics.
01
Upload Image
Submit any JPEG, PNG or WEBP image up to 20MB. Images are processed locally and never stored on our servers.
02
Dual Engine Analysis
The Deep Learning engine runs a SigLIP2 ensemble simultaneously with the CV engine's 11-signal forensic scan including ELA, DCT, SIFT, and JPEG ghost analysis.
03
Forensic Report
Receive a comprehensive report with per-signal scores, forgery type classification, confidence percentages, and EXIF metadata analysis.
Detection Capabilities
What We Detect
ForgeGuard classifies images into four distinct forgery categories, each with dedicated detection algorithms.
AI Generated
Fully synthetic images created by diffusion models, GANs, or vision models. Detected via ELA uniformity, DCT frequency analysis, and deep learning ensemble scoring.
ELADCT FREQSIGLIP2COLOR STATSMETADATA
Copy-Move Forgery
A region from within the same image is copied and pasted elsewhere to conceal or duplicate objects. Detected via SIFT keypoint matching and DCT block similarity.
SIFT+RANSACDCT BLOCKSHASH MATCH
Image Splicing
Content from a different source image is composited into the target. Noise inconsistency, chrominance variance, and JPEG ghost artifacts reveal the composite origin.
JPEG GHOSTNOISE MAPILLUMINATIONEDGE SEAM
Inpainting / Removal
Objects erased and background synthetically filled using AI inpainting. Detected via texture repetition mapping, over-smooth region detection, and gradient discontinuity analysis.
TEXTURE REPOVER-SMOOTHGRAD DISC