FRAUD - Face Replay Attack UQ Datasets

SAS Logo



The digital representation of captured biometric signals may be injected into the system (replay attack) thus bypassing any liveness testing.

The FRAUD Datasets are the first practical attempt to counter replay attacks for face recognition systems that use uncontrolled consumer devices. The final operation of the system is designed to illuminate the face of the user with images displayed on the computer screen, and then using the camera to capture the face images that will be used for the face recognition system. In addition to liveness and anti-spoofing tests, the image reflection might be observed in the face.

If the reflection sequence matches the images that were displayed, it can be assumed that the video was created at the time that the images were displayed.

In the case of this dataset, the videos have been previously captured at the time the nominated images were displayed. In its final form, it is expected that the images will be nominated in real-time.

Researchers are invited to use these datasets to improve techniques for classifying image reflection from objects, and to validate the results obtained in our experiments.


The FRAUD1 dataset contains videos of objects using a computer tablet screen that displays different colours to illuminate them. The objects consist of a blank piece of white paper, a printed face on a piece of paper, and ten live subjects.


The FRAUD2 dataset contains videos recorded using three different webcams of objects that are illuminated by a computer tablet screen that displays only mono-chromatic (White) light. The objects consist of Soft Toys, printed faces on paper, and five live subjects.


Further enquiries can be made to:
{danny döt smith ät uq döt net döt au},
{a döt wiliem ät uq döt edu döt au}, or
{lovell ät itee döt uq döt edu döt au}.

Last updated: 04-Feb-2015