Blood flow detection: New weapon against deepfakes

Deepfakes, in which videos are digitally manipulated, pose a growing threat. The Netherlands Forensic Institute (NFI) has developed an innovative method to detect these fake videos. Investigators can determine whether images are authentic by monitoring subtle discolorations in the face caused by blood flow during a heartbeat. Deepfakes lack these small changes, which makes them detectable.

Zeno Geradts, the forensic researcher behind this development, discovered that this idea originated in research conducted by MIT in 2012. The technique is up-and-coming for use in criminal investigations, such as verifying evidence on camera footage. Despite the progress, Geradts is also concerned that we are entering an era where no image can be trusted.

Blood flow in image analysis

The basis of blood flow detection lies in measuring vascular changes around the eyes, forehead, and jaw, where blood vessels lie close to the skin. With each heartbeat, these veins expand, leading to a subtle but measurable color difference. This technique, originally derived from medical research into heart rate measurement via video, demonstrates that physiological signals can be reliable indicators of image authenticity. The NFI's approach emphasizes the need to look beyond pixel comparisons and focus on biological parameters that are difficult to falsify.

Deepfake fraud hits half of global businesses

49% of companies worldwide are victims of virtual crimes, new study reveals.

Read More

Deepfakes have evolved into high-quality videos that replicate real heartbeats. Innovations such as remote photoplethysmography (rPPG) have shown that heartbeat signals can be transferred from original images into deepfakes, making them significantly more challenging to detect. Fraunhofer HHI researchers have discovered that the latest deepfakes even retain the authentic physiological signatures of the original subjects. This development calls for new detection methods focusing on more detailed physiological data, such as local blood flow patterns. At the same time, cryptographic markers such as digital fingerprints may also become critical standards.

Challenges

Although current advances in blood flow detection are promising, challenges remain in detecting modified videos. Detection methods must evolve further as the latest deepfakes can mimic realistic heartbeats. Solutions such as digital fingerprints and more detailed analyses of physiological signals are currently being developed to distinguish deepfakes from authentic images. Advances in technology must keep pace with the increasing complexity of deepfakes to effectively combat deception and disinformation.

Blood flow detection offers a groundbreaking tool for forensic investigation. The NFI plans to implement this technology in criminal cases to ensure the integrity of visual evidence. In a world where images are increasingly manipulated, the legal world must have tools to confirm video material's authenticity and reliability. Zeno Geradts, who is further refining the blood flow detection method, will present his findings at a European conference at the end of this month, opening the door for broader application in international legal proceedings.