Perceptual hashing library in python (with redis), a wannabe PhotoDNA
Perceptual hashing is the use of an algorithm that produces a snippet or fingerprint of various forms of multimedia.[1][2] Perceptual hash functions are analogous if features of the multimedia are similar, whereas cryptographic hashing relies on the avalanche effect of a small change in input value creating a drastic change in output value. Perceptual hash functions are widely used in finding cases of online copyright infringement as well as in digital forensics because of the ability to have a correlation between hashes so similar data can be found (for instance with a differing watermark). Based on research at Northumbria University,[3] it can also be applied to simultaneously identify similar contents for video copy detection and detect malicious manipulations for video authentication. The system proposed performs better than current video hashing techniques in terms of both identification and authentication.
Pic Source: Why we created 'Imageid' and saved 47% of the moderation effort | by Diego Essaya | Taringa! | Medium
Perceptual hashing converts an image, by degrading it and turning it into "pixels", into a binary (or hexadecimal) sequence. Unlike cryptographic hashing, perceptual hashing lacks of avalanche effect, making any change in the image easily perceivable in the hash.
It uses phash and whash by checking initially phash, then whash.
By combining these two with a db (redis), you get this library.
You can:
- Ban images: Add the hash of the image to the DB (and checks if already in it). This includes rotations (90 degrees left right 180 up down) of the pictures.
- Unban images: Remove the hash and all the similar hashes from DB;
- Whitelist images: Ignore a picture hash.
Perceptual hashing is a good way to recognize two similar images. If you need to:
- Fast indexing similar images;
- Check for prohibited content without saving it into your DB (child pornography, pornography, porn, gore...);
- Check for watermarked original copyrighted content.
and more...
The library can easily detect an edited photo if it has:
- Color changes;
- Random garbage over it (watermarks, stickers....);
- slight cropping.
Remember that this is not ML-Based.
It can be easily bypassed by cropping the image.
This library is a wannabe PhotoDNA.
-
Install redis
-
Start redis
-
git clone https://github.com/matteounitn/iHashDNA.git -
cd into folder
-
(Optional) create a venv:
python3 -m venv venv && source venv/bin/activate -
pip3 install -r requirements.txt
Then you are good to go!
Checkout this example.