Video is All You Need: Attacking PPG-based Biometric Authentication

Abstract

Unobservable physiological signals enhance biometric authentication systems. Photoplethysmography (PPG) signals are convenient owning to its ease of measurement and are usually well protected against remote adversaries in authentication. Any leaked PPG signals help adversaries compromise the biometric authentication systems, and the advent of remote PPG (rPPG) enables adversaries to acquire PPG signals through restoration. While potentially dangerous, rPPG-based attacks are overlooked because existing methods require the victim's PPG signals. This paper proposes a novel spoofing attack approach that uses the waveforms of rPPG signals extracted from video clips to fool the PPG-based biometric authentication. We develop a new PPG restoration model that does not require leaked PPG signals for adversarial attacks. Test results on state-of-art PPG-based biometric authentication show that the signals recovered through rPPG pose a severe threat to PPG-based biometric authentication.

Identification of PPG Measurement Sites Toward Countermeasures Against Biometric Presentation Attacks

“…In addition, as shown in Fig. 1, several studies have warned PAs may be performed on PPG-based biometric authentication using various PPG measurement sites on a victim's body and generating fake PPG signals to inject into the smartwatch as the authentication device [10], [16]. Seepers et al investigated a PA on PPG-based authentication using heartbeats estimated based on PPG recorded at the face using a camera [10].…”

“…We also investigated a PPG waveform mapping method to generate fake signals based on frequency characteristics to improve PA [18]. In addition, Li et al proposed a waveform mapping method using PPG signals recorded at the face and a generative adversarial network (GAN) for a PA on PPG-based authentication [16]. Therefore, countermeasures are required against PAs by using waveform mapping techniques for PPG-based authentication.…”

“…Meanwhile, PPG waveforms as biometric quality depend not only on individuals but also on measurement sites [11], [23]. PAs against PPG-based authentication are based on the leakage of the information required for authentication derived from PPG signals recorded at impostor measurement sites [16], [18]. Therefore, if liveness detection can be performed using one PPG signal as a biometric quality-based technique, it may connect PPG-based authentication and the countermeasure against the PA with one PPG sensor [7].…”

“…For example, the finger and wrist, and the earlobe and forehead provide more similar PPG waveforms based on the blood vessel configuration and blood circulation [11]. In addition, Li et al evaluated the feasibility of PA using PPG at the face using a camera, and concluded that the difficulty of PA is due to the camera specification such as frame rate when generating fake signals [16]. Therefore, the proposed countermeasure may identify PPG signals on the face using a camera with a low frame rate and mapped signals using the PPG signals on the face.…”

“…We applied only one technique using the TF computed using PPG signals from participants based on the frequency characteristics, although the result exhibited the effectiveness of estimating PPG wr . Techniques for mapping based on deep learning techniques, such as GAN [16], are available, which might generate fake signals that are more similar to genuine PPG signals compared with our technique.…”