Die Frequenzanalyse offenbart typische Artefakte in computergenierten Bildern.
© RUB, Marquard

Information Technology Recognising fake images using frequency analysis

This method exposes fake images created by computer algorithms rather than by humans.

They look deceptively real, but they are made by computers: so-called deep-fake images are generated by machine learning algorithms, and humans are pretty much unable to distinguish them from real photos. Researchers at the Horst Görtz Institute for IT Security at Ruhr-Universität Bochum and the Cluster of Excellence “Cyber Security in the Age of Large-Scale Adversaries” (Casa) have developed a new method for efficiently identifying deep-fake images. To this end, they analyse the objects in the frequency domain, an established signal processing technique.

The team presented their work at the International Conference on Machine Learning (ICML) on 15 July 2020, one of the leading conferences in the field of machine learning. Additionally, the researchers make their code freely available online, so that other groups can reproduce their results.

Interaction of two algorithms results in new images

Deep-fake images – a portmanteau word from “deep learning” for machine learning and “fake” – are generated with the help of computer models, so-called Generative Adversarial Networks, GANs for short. Two algorithms work together in these networks: the first algorithm creates random images based on certain input data. The second algorithm needs to decide whether the image is a fake or not. If the image is found to be a fake, the second algorithm gives the first algorithm the command to revise the image – until it no longer recognises it as a fake.

Ein Teil des Bochumer Forschungsteams: Thorsten Holz, Lea Schönherr, Joel Frank und Thorsten Eisenhofer (von links)
© RUB, Marquard

In recent years, this technique has helped make deep-fake images more and more authentic. On the website www.whichfaceisreal.com, users can check if they’re able to distinguish fakes from original photos. “In the era of fake news, it can be a problem if users don’t have the ability to distinguish computer-generated images from originals,” says Professor Thorsten Holz from the Chair for Systems Security.

For their analysis, the Bochum-based researchers used the data sets that also form the basis of the above-mentioned page “Which face is real”. In this interdisciplinary project, Joel Frank, Thorsten Eisenhofer and Professor Thorsten Holz from the Chair for Systems Security cooperated with Professor Asja Fischer from the Chair of Machine Learning as well as Lea Schönherr and Professor Dorothea Kolossa from the Chair of Digital Signal Processing.

Frequency analysis reveals typical artefacts

To date, deep-fake images have been analysed using complex statistical methods. The Bochum group chose a different approach by converting the images into the frequency domain using the discrete cosine transform. The generated image is thus expressed as the sum of many different cosine functions. Natural images consist mainly of low-frequency functions.

In den Frequenzbereich transformierte Personenbilder: Die linke obere Ecke repräsentiert niederfrequente Bildbereiche, die rechte untere Ecke hochfrequente Bereiche. Links ist die Transformation eines realen Personenfotos zu sehen, der Frequenzbereich ist gleichmäßig verteilt. Die Transformation des computergenerierten Fotos (rechts) enthält eine charakteristische Gitterstruktur im hochfrequenten Bereich – ein typisches Artefakt.
© RUB, Lehrstuhl für Systemsicherheit

The analysis has shown that images generated by GANs exhibit artefacts in the high-frequency range. For example, a typical grid structure emerges in the frequency representation of fake images. “Our experiments showed that these artefacts do not only occur in GAN generated images. They are a structural problem of all deep learning algorithms,” explains Joel Frank from the Chair for Systems Security. “We assume that the artefacts described in our study will always tell us whether the image is a deep-fake image created by machine learning,” adds Frank. “Frequency analysis is therefore an effective way to automatically recognise computer-generated images.”

Original publication

Joel Frank, Thorsten Eisenhofer, Lea Schönherr, Asja Fischer, Dorothea Kolossa, Thorsten Holz: Leveraging frequency analysis for deep fake image recognition, International Conference on Machine Learning (ICML), 2020, virtual conference, online publication

Press contact

Joel Frank
Chair for Systems Security
Faculty of Electrical Engeneering and Information Technology
Ruhr-Universität Bochum
Germany
Phone: +49 234 32 27814 (Secretary's office)
Email: joel.frank@rub.de

Download high-resolution images
The selected images are downloaded as a ZIP file. The captions and image credits are available in the HTML file after unzipping.
Conditions of use
The images are free to use for members of the press, provided the relevant copyright notice is included. The images may be used solely for press coverage of Ruhr-Universität Bochum that relates solely to the contents of the article that includes the link for the image download. By downloading the images, you receive a simple right of use for one-time reporting. Saving the images for other purposes or further processing of the images that goes beyond adapting them to the respective layout requires an extended right of use. Should you therefore wish to use the photos in any other way, please contact redaktion@ruhr-uni-bochum.de

Published

Thursday
16 July 2020
9:03 am

By

Julia Weiler (jwe)
Christina Scholten

Translated by

Donata Zuber

Share