IdeaWell

Project ideas from Hacker News discussions.

ZK-JPEG: Zero-Knowledge Image Editing and Compression

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1 – Technical capability & the need to scrutinize transformations
The ZK‑JPEG approach can prove that an image has undergone only certain allowed edits, but users must still examine those edits to judge whether they look legitimate.
- “our tool can verify a large family of image transformations… Is this family large enough to transform real image A into arbitrary fake image B?” – Retr0id
- “Seems like the answer to my question is 'yes', so you have to carefully inspect the transformations to see if they look legit.” – Retr0id
- “If you're preserving the original, then you don't need the ZKP… you could apply redactions or make the file smaller, without invalidating the signature.” – Retr0id

Theme 2 – Defining acceptable operations / policy layer
Beyond proving a transformation is possible, a practical system needs a clear policy of which edits are permitted (e.g., compression, color‑correction, redactions).
- “Presumably in a real application there would be a list of acceptable operations… Step 2 is coming up with a policy of what restrictions to place on allowed transformations.” – bawolff
- “Hasn't this crossed over into being a philosophical question? You want to attest to reality or provenance… then you start making lists of 'acceptable' changes… It then slowly slides into 'average human perceptibility'; 'irrelevant details'…” – jamienk

Theme 3 – Philosophical & social implications for image trust in the AI era
The discussion questions whether we can restore pre‑AI levels of photo reliability, probing intent, perception, and the cultural weight of images.
- “We are trying to make images in the AI era be as reliable as they were pre‑AI? But they were not reliable pre‑AI, it was just more difficult to intentionally doctor them.” – jamienk
- “Tie this into the Apple/Android/Sony/Leica signed photos, and you get provenance from capture to publish.” – mvid
- “I'm trying to think of recent real‑world examples where the reality of the photo was of major importance… Photographs just don't seem to have the same significance…” – AmazingEveryDay

🚀 Project Ideas

ProvablePic

Summary

  • A service that signs raw sensor data (or a trusted original) and generates zero‑knowledge proofs for allowed image transformations (crop, redaction, color‑correction, compression) so any derivative can be verified as originating from the attested source.
  • Core value proposition: publishers and platforms can display a cryptographically verified “real” badge while still permitting necessary edits, addressing the trust gap highlighted by HN commenters.

Details

Key Value
Target Audience Photojournalists, news platforms, real‑estate marketplaces, social media sites
Core Feature ZK proof generation/verification for a policy‑defined set of image transformations, bound to a signed original
Tech Stack Rust (circuit generation with halo2), WebAssembly for client‑side verification, IPFS/FileStorage for originals, API in Go/Node
Difficulty Medium
Monetization Revenue‑ready: tiered API pricing per proof verification (free tier, paid for high volume)

Notes

  • HN users asked for “a list of acceptable operations” (bawolff) and wanted to “preserve the original” while allowing redactions (Retr0id) – ProvablePic lets platforms enforce such policies cryptographically.
  • Provides a practical counter to the “analog gap” attack by tying the proof to the original sensor data, making re‑photographing a printed fake ineffective without breaking the proof chain.

EditTrace

Summary

  • An open‑source tool that records an immutable edit manifest (ordered list of operations) alongside an image, enabling viewers to step through each transformation and see the original at any point.
  • Core value proposition: gives audiences the transparency to verify how an image was derived, directly addressing the desire to “see the original before transforms” and “apply each one at a time” (gblargg, Retr0id).

Details

Key Value
Target Audience Editors, journalists, content creators, forensic analysts
Core Feature Append‑only JSON manifest (operation type + parameters) stored in a tamper‑evident log (e.g., Merkle tree or blockchain) with a viewer UI that replays edits
Tech Stack TypeScript/React for UI, Electron for desktop, SQLite + Merkle‑tree backend, optional IPFS for asset storage
Difficulty Low
Monetization Hobby

Notes

  • Commenters like gblargg explicitly requested a viewer that shows “the original before transforms, and perhaps apply each one at a time”; EditTrace fulfills that need.
  • Enables discussion around what transformations should be considered acceptable, as the manifest makes the edit policy explicit and auditable.

AuthentiMark

Summary

  • A steganographic watermarking system that embeds a cryptographic commitment to the original image (and optionally AI‑generation metadata) into the image’s noise‑resistant bits, allowing detection of unauthorized AI‑generated layers or deepfakes.
  • Core value proposition: provides a lightweight, invisible provenance marker that survives common compressions and resizing, helping platforms flag AI‑doctored content as requested by HN users concerned about “AI faking photos” (Gigachad, avianlyric).

Details

Key Value
Target Audience Social media platforms, news agencies, digital asset marketplaces
Core Feature Embed/extract a verifiable commitment using error‑correcting codes; detection API flags mismatched or missing marks
Tech Stack Python (numpy, pywt for wavelet domain), Rust for high‑speed extraction, REST API, Docker deployment
Difficulty Medium
Monetization Revenue‑ready: per‑month subscription based on number of images processed

Notes

  • HN discussion highlighted the need for a “tick showing the photo is verified real” to add trust to marketplaces (Gigachad) and worries about AI‑generated images in politics (avianlyric); AuthentiMark gives that verifiable signal.
  • Provides a practical tool for journalists to prove they haven’t inserted AI‑generated layers, echoing the desire to “minimize the dishonesty of using generative AI” (Retr0id).

Read Later