Detection of Face Spoofing Using Visual Dynamics

Abstract

Abstract-Rendering a face recognition system robust is vital in order to safeguard it against spoof attacks carried out by using printed pictures of a victim (also known as print attack) or a replayed video of the person (replay attack). A key property in distinguishing a live, valid access from printed media or replayed videos is by exploiting the information dynamics of the video content, such as blinking eyes, moving lips, and facial dynamics. We advance the state of the art in facial anti-spoofing by applying a recently developed algorithm called Dynamic Mode Decomposition (DMD) as a general-purpose, entirely data-driven approach to capture the above liveness cues. We propose a classification pipeline consisting of DMD, Local Binary Patterns (LBP), and Support Vector Machines (SVM) with a histogram intersection kernel. A unique property of DMD is its ability to conveniently represent the temporal information of the entire video as a single image with the same dimensions as those images contained in the video. The pipeline of DMD+LBP+SVM proves to be efficient, convenient to use, and effective. In fact only the spatial configuration for LBP needs to be tuned. The effectiveness of the methodology was demonstrated using three publicly available databases: print-attack, replay-attack, and CASIA-FASD, attaining comparable results with the state of the art, following the respective published experimental protocols.

Movement correction in DCE-MRI through windowed and reconstruction dynamic mode decomposition

“…These modes essentially capture different large-scale to smallscale structures (sparse components) including a background structure (low-rank model) [7]. DMD has gained significant applications in various fields [2,3,16], including for detecting spoof samples from facial authentication video data sets [33] and for detecting spoofed finger-vein images [31]. The advantage of this method is its ability to identify regions of dominant motion in an image sequence in a completely data-driven manner without relying on any prior assumptions about the patterns of behaviour within the data.…”

“…To compensate the periodic free breathing from the DCE-MRI sequence in this study, we consider W = 3. For instance, running DMD on the window containing the first three images in the sequence, we obtain two dynamic modes (in general for N images, we get N − 1 DMD modes [33]). The first dynamic mode 'c1' captures the low-rank image across the window and the second dynamic mode 'c2' captures the sparse representation, which essentially contain motion artefacts pertaining to periodic free breathing.…”

One dimensional local binary patterns of electroencephalogram signals for detecting Alzheimer's disease

“…Throughout the experiments, we shall use EER to optimize the decision threshold and report performance in HTER [22], [23].…”

“…Contributions of the paper are as follows: Although LBP has been used extensively as a feature descriptor in the fields of image [16], [17], [22], [23], speech [18], [24] and signal processing [18], [25], its application to analysing AD patients' data is novel. We show that 1D-LBP captures the descriptive information, in terms of histograms representing the relative changes in EEG amplitudes, in a way that can be readily utilised by a support vector machine (SVM) for classification.…”

“…Although HTER has been used extensively in the field of biometrics [22], [23], its usage in biomedical applications is not very well known. HTER has the advantage of not being affected by the overwhelmingly large sample size of one class versus another because both types of errors are weighted equally, thus coercing equal contributions from both errors.…”