On December 27th, the 16th China Security Forum Technology Forum was held during the CPSE Security Expo. Tang Linpeng, co-founder and CTO of Moqi Technology, delivered a keynote speech at the forum on “AI+ Trusted Biometric Engine: Accurate, Reliable, Active and Convenient, and Protects Privacy”, which comprehensively explained the leadership of trusted biometric technologies driven by technological innovations such as AI. properties and broad application prospects.
From 2020 to 2025, the global biometrics industry will grow rapidly at a compound annual growth rate of 14.6% per year, and it is expected to reach nearly 70 billion US dollars by 2025*, and traditional fingerprint and palmprint biometrics will still occupy a large share of the market .
With the rapid development of 5G, artificial intelligence, AIoT, cloud computing, and digitalization, it will further promote the vigorous development of the biometric market. For this reason, the society needs high-precision, high-reliability, smooth and seamless, and privacy-protecting biometric technologies and products. , to allow people to move safely and easily between the physical and digital worlds.
However, there are also hidden concerns under the prosperity – personal biometrics, due to their life-long properties, will cause greater security risks once leaked, and unlike passwords, once leaked, it cannot be changed or recovered. Therefore, the privacy and security issues of biometric technology and data are extremely critical.
Three characteristics of next-generation trusted biometrics: accurate and reliable, active and convenient, and privacy protection
On November 1 this year, the “Personal Information Protection Law of the People’s Republic of China” was officially implemented. The law clarified the principle of informed consent and the principle of minimum necessity for personal information processing, and also required technical means such as data desensitization, de-identification, and anonymization. Keep sensitive personal information safe.
In order to protect the important and sensitive personal information of “original biological features”, Moqi Technology proposed that the next generation of trusted biometrics should have three characteristics: First, it is accurate and reliable, that is, biometrics can be used in large storage capacity without a large amount of data labeling. High precision and high performance are achieved; second, it is active and convenient to achieve a natural, easy-to-use, informed and controllable interactive experience; the third is to protect privacy, which can be truly provable from the source technology to protect privacy.
Moqi Technology focuses on non-contact fingerprint and palm vein recognition. The biometric identification technology on these “hands” has advantages in accuracy, active convenience, and privacy protection.
For example, the Moqi M1 smart all-in-one machine recently launched by Moqi Technology adopts the dual-mode fusion technology of palm print + palm vein. Among them, palm print has multi-scale features, rich information and high precision; palm vein identification uses near-infrared imaging technology to image the blood vessels of the blood, with its own living body information, difficult to steal, preventing counterfeiting attacks, and high privacy; Non-contact collection and identification, easy to use, smooth experience, easy to use, high privacy and security.
The dexterous hands allow human beings to operate and control tools, making human beings intelligent creatures, and Moqi hopes to use technology to help people open the door to the digital universe in the AI era with both hands, making biometric identification more natural, friendly and controllable .
Moqi Technology’s intelligent and trusted biometrics engine is based on fingerprints, finger veins, palm prints, palm veins and other biometric features on the hand as the origin, supported by AI underlying technology innovation, with safe and convenient non-contact interaction as the contact, empowering smart security , smart city, finance, cultural tourism, immigration, pension and other government and enterprise affairs, leading biometrics to enter a new stage of accuracy, reliability, initiative and convenience, and privacy protection.
Accurate and Reliable: Entering the billion-level era of high-precision and high-reliability biometric identification
In terms of accuracy and reliability, Moqi Technology can achieve the only certification of billions of levels of high precision and high performance. In the 1:1 verification scenario, the accuracy is one in 50,000. In the 1:N recognition, when the scale of N reaches 1 billion, this accuracy represents 20,000 errors and is completely unavailable. The algorithm has put forward very high requirements.
However, the current mainstream deep learning-based image search solutions have some limitations. For example, they rely on massive training data and are sensitive to some complex environments such as occlusion, illumination, and perspective transformation. They need to build a more complex infrastructure to deal with massive amounts of data. In practice, image search results are often inaccurate.
To this end, Moqi Technology has developed a new set of unstructured data representation and retrieval technologies. With the representation of multi-scale vectors and graphs, the image is represented by vectors and graphs at each layer from the whole to the part. The core multi-scale characterization enables the feature information to achieve multiple orders of magnitude improvement over a single-scale vector.
At the same time, Moqi will also establish a search-friendly representation based on multi-scale representations, such as multi-stage, multi-strategy fusion heterogeneous retrieval, to achieve high-speed and high-precision results. In terms of training data requirements, because of the realization of multi-scale image representation at the bottom layer, Moqi can greatly improve the utilization of a single image, and adopts a self-supervised learning method. Compared with only relying on end-to-end deep learning, the amount of training data required has been reduced by an order of magnitude, realizing Moqi’s small-sample self-learning framework, which enables Moqi Technology to break through the large library attenuation problem of traditional biometrics. For low-quality fingerprints, with a storage capacity of 1 billion, the top ranking rate reaches 90%, and the top five ranking rate reaches 99%, and no manual labeling is required, which greatly improves the efficiency and helps users take the lead in achieving 1 billion levels in the world. Second-level, high-precision, automated search and recognition on images.
At present, Moqi Technology has supported the construction of a smart palmprint platform with a level of over 4 billion, supports real-time comparison on billions of palmprint images, and has high precision, high performance, intelligence and security. , easy maintenance and other features, the service has been applied to 28 provinces across the country.
Active Convenience: Achieve an active and informed contactless convenience experience
Traditional fingerprint collection and identification often require contact, especially in the case of an epidemic, there is a risk of infection.
Moqi Technology uses multi-eye vision and structured light, 3D reconstruction, image enhancement and other AI technologies to fully upgrade the traditional contact-based fingerprint technology to the non-contact era. Users only need to reach out to complete the collection, identification and temperature measurement, and the experience is smooth, efficient and convenient. At the same time, the whole process requires user authorization, information is controllable, and privacy security is more protected.
Protecting Privacy: Creating an Indestructible “Safe” for Bioinformatics
Moqi Technology believes that the next generation of biometric technology should protect privacy at the source. At present, the mainstream application of biological information is to retain the original image information or save the template on the platform. Even the template can still recover quite a few original biological features, which cannot effectively protect privacy.
How to use biometrics and protect privacy? This requires that the biometric platform does not store the original biometric, or a template of the original biometric, but uses a transformed biometric template. Moqi Technology deeply integrates and innovates biometrics and the underlying cryptographic tools, so that the transformed template satisfies the three characteristics of irreversibility, revocation and non-association. Similar to a safe, only real users can open the safe to get the key and unlock the app.
In terms of technical implementation, Moqi solves the inconsistency of features caused by changes in the acquisition environment through error correction codes, and also improves the fault tolerance rate. In order to further protect privacy and security, Mo Qi added a lot of noise to the biometrics to protect the original biometrics from being recovered. What the attacker sees is only chaotic, true and false feature points. The cracking work is equivalent to solving an NP- Hard math problems to protect privacy and mathematically prove its security.
The next-generation credible biometric identification technology created by Moqi Technology is accurate and reliable, and can achieve high-precision and high-performance biometric identification under billions of storage capacity; at the same time, it has the characteristics of active and convenient, and users can reach out and know the user. Controllable natural interaction; in addition, Moqi’s biometric technology fully protects privacy, and achieves strong privacy security that can be mathematically proven at the source.
Combining these three characteristics, Moqi Technology has also launched products and solutions for different industries and application scenarios, opening up digital and physical identities, linking the identities of the online world and the physical world, enabling travel, cultural tourism, hotels , community, payment, hospital, pension and other industries, providing a more convenient and safer identity authentication experience for the industry.
*Data from third-party research institutions