The front foot has just begun to try fingerprint payment, and the back foot has entered the face brushing era. With the technology upgrading, biometric technology will be applied to all walks of life in the future, penetrating every aspect of people’s life. Your face, eyes, gait and even expression are telling the recognition system who you are.
High risk of single biometric identification
However, biometrics is not foolproof. Idiap Research Institute has studied the detection effect of current face recognition technology on deepfake (an AI deep face changing software). The results show that the current face recognition models and detection methods are basically helpless in the face of deepfake: the image classification model VGg with excellent performance and the algorithm based on facenet have an error rate of 95% when distinguishing true and false videos; The lip shape based detection method can not detect whether the speech and mouth shape of the characters in the video are consistent.
Of course, not only face recognition can be cracked. Biometric recognition methods such as fingerprint recognition, iris recognition and vein recognition all have cracking methods.
Multimodal identification can combat higher risks
Does this indicate that biometrics is unsafe? The answer is No. Biometrics of different modes have their own characteristics and different anti forgery abilities. People are born with many “sensors” – ears, nose, eyes, hands, and other senses. The ultimate goal of artificial intelligence is to do everything like people, and these sensors are also needed. To combat the higher level of prosthesis attack risk, multimodal fusion can be used.
For example, gait and face integrated recognition, face recognition conditions are relatively high, and gait can be clearly imaged even in long-distance complex scenes. Therefore, face decision weight and gait decision weight can be combined, which can not only greatly increase the difficulty of data forgery, but also improve the recognition security.
Finance and access control market are the entry points
At present, fingerprint and finger vein composite recognition, face and fingerprint composite recognition and other technologies have emerged in the market. Biometric manufacturers can not only consolidate their market segmentation advantages, but also appropriately develop new composite technologies, or cooperate with other manufacturers to win greater market share with composite biometric products.
In many application fields, the financial industry has a great demand for multimodal identification. Multimodal biometric technology helps to reduce the losses caused by financial fraud and ensure the security of user information or data. It is better than single biometric technology, or it will become the future trend of financial technology. However, the total number of branches and outlets of the bank across the country is 100000, and the overall planning and preparation work, such as product form planning and background service system construction, is huge, so it takes some time to promote and implement.
In addition, it is worth noting that there is a large demand for updating and iterating in the smart community equipment stock market at present. Face recognition equipment is used in many community access control and gateways. However, with the increasing scale of face recognition equipment, its weaknesses will be gradually exposed, and multimodal equipment combined with other recognition technologies will break the limitations of simple face recognition products and become the mainstream products in the access control market. Smart community access control and gate will become an emerging potential market for multimodal identification.
Conclusion: there is no doubt that in today’s increasingly severe security situation, multimodal identification will be the development trend of biometrics in the future. However, there is no absolutely safe technology, and systems that are even safer in design may have vulnerabilities due to human or accidental factors. However, this does not mean that these technologies cannot be applied. Instead, the alarm should be kept ringing to better control risks, strictly follow standards, and strive to reduce risks to a low level.