PleasEye 01 / 14
Çankaya University · SENG 491-492

PleasEye

Proving photos are real, the moment they are taken. Fighting fakes with math, not guessing.

Graduation Project, 2025-2026 Software Engineering
Live 02 / 14
A small experiment

Which one is real?

One was filmed by a real camera. The other was made by AI. Hit reveal to find out which is which.

Video A
Video A drop file at
presentation/assets/fake.mp4
Video B
Video B drop file at
presentation/assets/real.mp4
Guessing is not enough PleasEye gives proof
The Problem 03 / 14
The Deepfake Problem

What you see may not be real.

AI can now create images that look completely real. Tools that try to catch fakes are always one step behind.

  • 01 Fake images everywhere AI creates realistic fakes in seconds.
  • 02 Nobody knows what to trust News, courts, and social media cannot tell real from fake.
  • 03 Fake-catchers fall behind Every new AI model beats the old detectors.
LaurieWired tweet about proof of physical capture
@lauriewired 1.4M views
The old wayAlways catching up
Solution 04 / 14
Our Approach

Sign at capture. Verify anywhere.

Instead of checking if a photo is fake later, we sign it the moment it is taken.

The flow
01 Take the photocamera creates a unique fingerprint of the image
02 Sign ita secure chip stamps it with a secret key
03 Build a recordfingerprint + time + device info
04 Save foreverwritten to a public record that nobody can change
What we prove
  • AWho took itwhich device captured the photo
  • BNothing changedthe photo has not been edited
  • CWhenthe exact time of capture
  • DDevice is safeno one tampered with the hardware
Signed at the sourceStored forever
Architecture 05 / 14
How It Works

From camera to public proof, one path.

Three layers: the camera, the server, and the public record. Data flows in one direction only.

  • FWCamera softwaretakes photo, signs it, encrypts it
  • APIVerification serverchecks the signature, blocks reuse
  • L1Public recordstores proofs nobody can change
ESP32-CAMtakes and signs the photo L1
AES-256-GCMencrypts the package L2
Express APIchecks and verifies L3
Blockchainproof that cannot be changed L4
Camera, Server, ChainOne direction only
The Capture 06 / 14
A Real Photo

This is what gets signed.

A real photo from our camera. Low quality on purpose. The goal is proof, not pretty pictures.

  • DEVESP32-DEV-01which camera took it
  • SENOV2640 CMOS640 × 480 pixels
  • SEQ#0042photo number (goes up by one each time)
  • SHAa1b2c3d4 … 9f8e7d6cunique fingerprint of this photo
Frame · ESP32-DEV-01 · seq 0042 640 × 480
Real ESP32-CAM capture
signed at capture SHA-256 · a1b2c3…
Real output from the deviceSigned on the camera itself
Cryptography 07 / 14
The Security Steps

Hash. Sign. Encrypt. Store.

Five clear steps. No AI involved. Just standard math that anyone can check.

01 Take the photoraw image from the camera sensor
02 Create fingerprinta unique code for this exact image
03 Sign with secret keyonly this device can make this signature
04 Encrypt for transportprotected while being sent to the server
05 Build the proofone signed file with all the info above
// manifest.json { "aad": { "protocol_version": "pe-v1", "device_id": "ESP32-DEV-01", "seq": 42, "capture_timestamp": "...", "hash_algorithm": "SHA-256", "image_hash": "A1B2C3...", "tamper_state": "OK" }, "encryption": { "algorithm": "AES-256-GCM", "nonce": "base64...", "tag": "base64..." }, "signature": { "algorithm": "ECDSA-P256", "sig": "base64..." } }
No AI neededAnyone can check
Demo 1 drop file at
presentation/assets/demo1.mp4
Demo 2 drop file at
presentation/assets/demo2.mp4
Live Verification drop file at
presentation/assets/verification-demo.mp4
Live verification
Stack 10 / 14
Our Tools

What we built it with.

Simple stack. Standard security. Public record.

Hardware

  • ESP32-CAMcapture
  • SE050secure element
  • mbedTLScrypto lib

Backend

  • ExpressAPI
  • TypeScripttypes
  • Node cryptoverify

Ledger

  • Suipublic blockchain
  • Movesmart contracts
  • @mysten/suiSDK

Frontend

  • ReactUI
  • Vitebuild
  • Tailwindstyling
One clear pathCamera to public record
Features 11 / 14
Built-In Security

Safe by design.

Every part of the system has its own protection. If something goes wrong, the system shows it.

F1
Key stays on chip

The secret key never leaves the secure chip.

F2
Tamper alert

If someone opens the device, the key is erased.

F3
No reuse

Each photo gets a number. You cannot use the same proof twice.

F4
Cannot be changed

Once saved to the public record, nobody can edit it.

PropertyTraditionalPleasEye
Proves where the photo came from
Protected by hardware
Record cannot be changed
Works without AI
Hard to break into
SimpleStrong
Applications 12 / 14
Where It Matters

When a photo becomes proof.

One technology, many places where trust is important.

  • 01Important contentwhere knowing the source matters
  • 02Works everywheresame proof, many different uses
U1, News

Journalism

Check that news photos are real and not edited.

U2, Forensics

Legal Evidence

Prove a photo was not changed, from camera to courtroom.

U3, Verified

Social Media

Show a "verified real" badge on social posts.

U4, Compliance

Medical

Make sure medical and research images are real.

One technologyMany uses
The Vision 13 / 14
The Bigger Picture

A social feed where everything is real.

Every post has proof that it came from a real camera. No more guessing. Trust is the default.

@journalist Verified capture
Photo from the field, signed the moment it was taken.
@witness_lab Verified capture
Lab photo, signed by the camera, saved on the public record.
@news_wire Verified capture
News photo with full proof of where it came from.
@unknown Unsigned
No proof. Hidden by default.
No more guessingTrust is the default
Team 14 / 14
The Team

Four builders.

Software Engineering students at Çankaya University, building a graduation project on the future of digital trust.

EC
Efe Çelik
Software Engineer
BB
Batuhan Bayazıt
Software Engineer
BC
Barış Cem Bayburtlu
Software Engineer
BA
Burak Aydoğmuş
Software Engineer

Çankaya University

Software Engineering Department

SENG 491-492, 2025-2026

The PleasEye Team
The PleasEye Team 2026
←→ navigate