Biometric Authentication System Using Facial Data

Description

1 / 8 “BIOMETRIC AUTHENTICATION SYSTEM USING FACIAL DATA” Certificate of addition of invention BR 102022014438-9 dated 07 / 21 / 2022. Introduction

[001] This certificate refers to a complementary application, a complementary tool and complementary features applied to the already defined system of the invention above.

[002] As a complementary application, it deals with the integration of the possibility of carrying out fast Pix-type transfers through this same facial authentication system. This application will be incorporated into the system and will integrate the same facial registration, but will be applied to carrying out fast Pix-type transfers, which will increase security while maintaining the ease and speed of the operation.

[003] As a complementary tool, it deals with the incorporation of artificial intelligence into the facial analysis system. This tool aims at both general facial recognition already validated by the system and database checks to verify liveness, as well as more advanced analyses of realism and color and temperature patterns on the user's face. This aims to circumvent the fraud currently seen in similar systems, in order to increase security in carrying out important banking operations, such as Pix.

[004] As complementary features, it addresses the possibility of registration and access by legal entities, which is widely applied in national and international banks. This allows the registration of a person responsible for the transactions. Petition 870260017500, dated 25 / 02 / 2026, page 6 / 13 2 / 8 of a company's bank accounts, in addition to enabling the facial registration of two other representatives.

[005] Another complementary feature is the forced facial verification in pre-programmed payments, which aims to allow the scheduling of future payments, but requires facial recognition to be validated at the time of payment.

[006] Another complementary feature is the application of the payer's facial image to the recipient's transaction receipt. This allows for precise and instantaneous recognition of the payer, avoiding difficult-to-identify text, and provides a visual tool for the recipient to identify the payer and / or identify potential fraud. State of the art

[007] Brazilians are well aware that Pix is ​​an instant payment system created by the Central Bank of Brazil, launched in November 2020, that allows sending or receiving money 24 hours a day, seven days a week, with settlement in seconds and virtually no cost for individuals. By using "keys" (CPF, email, phone number, QR Code or NFC) linked to the account, it drastically simplifies the user experience and eliminates instruments such as TED, DOC and even in-person payment slips. Real-time settlement, combined with strict security rules and the mandatory participation of banks, fintechs and digital wallets, has led to unprecedented adoption: today Pix processes hundreds of millions of transactions daily and is already the most used payment method in the country. This scale reduces operational costs for companies, expands financial inclusion for millions of Brazilians without credit cards and enables new services — from collections Petition 870260017500, dated 25 / 02 / 2026, page 7 / 13 3 / 8 recurring tax payments — making it a critical infrastructure for the national digital economy.

[008] However, it was also observed that, although the core framework of Pix is ​​technically sound, daily practice has exposed gaps that deserve review. Social engineering scams still circumvent nighttime restrictions and precautionary blocking. In addition, 91% of refund requests made through the Special Return Mechanism end up being denied. For this reason, the Central Bank has already mandated mandatory scam alerts and improvements to the MED self-service by 2025, but the challenge of balancing convenience, speed, and user protection persists.

[009] It is also known that facial recognition authentication systems still have structural weaknesses. The most exploited is the presentation attack: printed photos, looped videos, or 3D masks can fool cameras that only capture RGB images when there is no robust liveness detection, especially if the software relies only on blinking or turning the head to prove "life." The popularization of deepfakes and direct video injections into the camera driver exacerbates the problem: in 2024, providers recorded an increase of up to 783% in attacks that replace the webcam stream with synthetic images in real time. Furthermore, since biometric models and templates are immutable, their leakage creates a permanent risk of usurpation. [0 1 0] . Taken together, these flaws show that facial authentication, while convenient, still needs complementary layers—multimodal sensors, template cryptography, and richer liveness proofs—to achieve security levels comparable to those of traditional multifactor methods. Petition 870260017500, dated 25 / 02 / 2026, page 8 / 13 4 / 8

[011] Finally, it is known that transaction receipts, especially for fast transactions like Pix, often present written information that is usually summarized or provides only superficial data about the payer. This text-based identification is flawed and often causes confusion when precise identification of the payer is required. Objectives of addition

[012] The addition of the complementary application to authorize Pix-type transactions through the facial recognition system proposed by the inventor aims to increase security by linking the transaction to a unique physical characteristic of the account holder, making it more difficult to take over accounts via SIM-swap or password leaks, reducing authenticity disputes and maintaining a more fluid "password-free" payment experience than tokens or OTPs. This complementary application is possible from the proposed system, as it does not alter the core infrastructure of Pix systems, but only adds verification for this type of transaction, which can be imported from the biometric (facial) verification already applied to access smartphone applications implemented by most national banks.

[013] The addition of the complementary trained artificial intelligence tool aims to elevate facial recognition to levels not yet achieved. The trained model is capable of identifying skin microtextures, lighting variations, and depth discrepancies impossible to reproduce in printed photos or "replay" videos. The AI ​​is also trained to perform a silent scan ("passive liveness") in external death databases to flag accounts whose holder has died, reducing the risk of takeovers. Petition 870260017500, dated 25 / 02 / 2026, page 9 / 13 5 / 8 account and maintaining frictionless KYC compliance for live users. This AI addition provides the system with an advanced barrier against fraud.

[014] The addition also aims to expand the usage and application features of the proposed system. This is done with: the possibility of registration and access by legal entities, which expands their possibilities within banking systems; the forced facial verification in pre-programmed payments, which aims to allow the scheduling of future payments, allowing better integration with banking systems; and the application of the payer's facial image on the transaction receipt, which allows the recipient to personally verify their identity and provides a quick visual identifier, eliminating the dependence on complex texts with poorly identifying information. Detailed description

[015] This addition defines the incorporation of an application into the biometric authentication system to allow authentication also in fast banking operations such as Pix.

[016] This application is based on the common condition of registering the account holder's face data for later recognition in validating an event or operation - in this case, authentication when making a payment or a monetary transfer via Pix.

[017] In this context, as an initiative for this addition to the system, it is noted that the accuracy rate of facial recognition for validation is 99.98% and that the facial data analyzed does not fade with age and cannot be modified or bypassed with silicone molds. Petition 870260017500, dated 25 / 02 / 2026, page 10 / 13 6 / 8

[018] This addition also provides for the incorporation of an additional tool into the original system, integrating artificial intelligence trained in the analysis of facial biometric data. This AI is trained to allow the verification steps foreseen in the original system and, in parallel, to distinguish a real face from a photograph, analyzing skin microtextures, lighting variations and depth discrepancies, preventing the use of photographs or virtual models as a form of fraud. It is also trained for passive liveness verification in general databases, preventing facial validation of deceased persons – this being a highly sophisticated anti-spoofing technology. Thus, the tool prevents authentication of users who use photographs, virtual models, silicone masks and the like, making real-time recognition or selfie taken at the moment necessary, where the AI ​​performs liveness tests to validate the data.

[019] In this scenario, this system and its digital tools use software and the internet for analysis and authentication, meaning it does not require special hardware attached to camera devices and eliminates any active user action to prove authenticity and liveness, only requiring the user to show their face to a device with an integrated camera. Furthermore, it does not require modifications to the 'card machines' or smartPOS, as the presence of the front camera is sufficient for authenticating the transaction by reading and analyzing the payer's facial data.

[020] The addition also brings a new feature, that of allowing registration and authentication for legal entities. In this registration, the facial data of the person responsible for payments in the company is used, but registration is allowed Petition 870260017500, dated 25 / 02 / 2026, page 11 / 13 7 / 8 facial recognition for up to two additional representatives to carry out banking transactions via facial recognition. Also, in the case of legal entities, this feature allows the facial registration process to be dynamic, meaning it can be updated whenever the individuals responsible for transactions within the registered company change.

[021] The addition also brings to the system the possibility of authentication in scheduled or recurring payments such as monthly installments or pre-dated payments and monthly expenses such as water, electricity, internet, etc. This scheduling of subsequent payments is carried out by the user / responsible party and authenticated by the system's facial recognition. In addition, for each pre-scheduled charge, the payment or debit is authorized by requesting facial recognition, where the system sends reminders before the due dates via the user's / responsible party's registered email and / or WhatsApp. This restricts automated payments, but provides security and double verification for their authorization in a necessary, yet easy and quick, facial recognition action.

[022] The addition also provides that, in this system, the Pix recipient will have the payer automatically registered in their list through the payer's face data identified in the payment history.

[023] The addition also provides that the system will print on the transaction receipts the selfie of the person who made the payment, that is, the image used at the time of biometric authentication. This gives the recipient a new visual identification of 'payment made' with the facial recognition of the person who Petition 870260017500, dated 25 / 02 / 2026, page 12 / 13 8 / 8 carried out the Pix or other transaction, replacing the written and often confusing texts currently available for verification by the recipient. Furthermore, it allows the recipient to verify the payer's authenticity and identify potential fraud.

[024] Thanks to the high security of the operation through facial recognition and advanced authenticity verification, the system eliminates value limits on remittances, allowing high-value and / or recurring transactions without the need for a user request. Optionally, this request can be implemented or chosen at the user's request.

[025] The addition also enhances the integration of the system into e-commerce, where the customer / user self-registers and authenticates their shipments with facial data, further facilitating online payments and preventing scams or undue charges. This, combined with the buyer's photo being stamped on the receipt, reduces to zero the chances of chargebacks due to fraud – a situation in which the holder of a credit or debit card disputes a purchase claiming that it was made without their authorization.

[026] This system allows all payment transactions using Pix or credit and debit cards to have facial authentication. Therefore, in addition to security for users, this solution assists in the KYC {know your customer} process against fraud, providing security and authenticated information to banks about the movement, origin and destination of funds, serving as a tool in cracking down on accounts of tax evasion, organized crime, trafficking and corruption, protecting the overall banking system. Funds without proven origin will be subject to investigation. Petition 870260017500, dated 25 / 02 / 2026, page 13 / 13

Claims

1 / 2 CLAIMS 1. “FACIAL DATA BIOMETRIC AUTHENTICATION SYSTEM” whose system, encompassing two processing phases: facial detection and authentication, is characterized by the fact that: it incorporates facial data biometric authentication in fast banking operations such as Pix; it integrates artificial intelligence trained in the analysis of facial biometric data, configured to allow the verification steps foreseen in the original system and, in parallel, to distinguish a real face from a photograph, analyzing skin microtextures, lighting variations and depth discrepancies, preventing the use of photographs or virtual models for authentication, and is also trained for passive liveness verification in general databases and to prevent facial validation of deceased persons.

2. “BIOMETRIC AUTHENTICATION SYSTEM USING FACIAL DATA” according to claim 1, characterized by the fact that it allows registration and authentication for legal entities, using the facial data of the person responsible for payments in the company, and allowing the facial registration of up to two additional representatives.

3. “BIOMETRIC AUTHENTICATION SYSTEM BY FACIAL DATA” according to claim 1, characterized by the fact that it allows authentication in scheduled or recurring payments such as monthly installments or pre-dated payments and monthly expenses such as water, electricity and internet, this scheduling of subsequent payments being carried out by the user / responsible party and authenticated by the system's facial recognition, wherein in each pre-scheduled charge the payment or debit is authorized by means of a facial recognition request, where the system sends reminders before the due dates via the user / responsible party's registered email and / or WhatsApp.

4. “BIOMETRIC AUTHENTICATION SYSTEM BY FACIAL DATA” according to claim 1, characterized by the fact that the Pix recipient will have the payer automatically registered in their list through the payer's facial data identified in the payment history.

5. “FACIAL DATA BIOMETRIC AUTHENTICATION SYSTEM” according to claim 1, characterized by the fact that it prints on bank transaction receipts the selfie of the person who made the payment, that is, the image used at the time of biometric authentication. Petition 870250049240, dated 12 / 06 / 2025, pp. 15 / 16