サイバーリンク List of News Articles

In the latest "Age Estimation Evaluation Test" conducted by NIST, the AI facial recognition engine "FaceMe" ranked 5th in the world. 【What is the NIST FATE Age Estimation Evaluation Test?】 The NIST FATE Age Estimation Evaluation Test is one of the international benchmarks for facial analysis technology conducted by the National Institute of Standards and Technology (NIST) in the United States, which evaluates the age estimation accuracy of AI algorithms using a large-scale facial image dataset. 【Evaluation Results of FaceMe】 - World Ranking: 5th FaceMe ranked 5th among 20 global vendors, demonstrating top-class accuracy in an international competitive environment. - Mean Absolute Error (MAE): 2.87 years The average error in FaceMe's estimated age is 2.87 years, with only a 0.35-year difference from the top-ranked algorithm, achieving world-class accuracy. *MAE (Mean Absolute Error): Quantifies the error in estimated age (the smaller the number, the higher the accuracy). *NIST conducted the evaluation using 1 million facial images of individuals aged 18 to 30.

In the Hanwha Vision showroom, only registered individuals can enter through high-precision facial recognition by FaceMe Security, achieving a strict access control system with a high level of security. ▶ Background and Details CyberLink and Hanwha Vision formed a partnership in 2023 to promote the seamless integration of the AI facial recognition security system "FaceMe Security" with Hanwha Vision's video management system "Wisenet WAVE," aiming to provide a multifunctional and advanced security system. With the opening of the Hanwha Vision showroom, we have implemented an access control system that integrates Hanwha Vision's AI cameras with FaceMe Security, allowing visitors to experience the technologies of both companies firsthand. - Authentication Camera: Hanwha Vision AI IR Wall Mount Camera (5MP AI IR Wall Mount Camera) - Facial Recognition System: CyberLink FaceMe Security - Main Features: Real-time detection of registered individuals' faces and automatic opening and closing of doors.

Details of the FaceMe Security 8.3 Update < Revamped UX for Action & Alert Rule Settings > Previously, individual settings were required for each action even under the same rule. By completely overhauling the UI and UX, we have transformed the operation screen to allow for intuitive and efficient configuration. ▶ Key Update Points - Multiple actions (notifications, access control, VMS integration, etc.) can now be configured under a single rule. - The settings screen has been consolidated into a single page, allowing for an overview and editing of the entire configuration at a glance. < Enhanced Integration Features with Face Terminal (Facial Recognition Terminal) > FaceMe Security integrates with the MSK Knoctoi series (Standard/Lite), enabling door opening and closing based on facial recognition results without the need for additional devices. This update allows for detailed configuration of door opening and closing rules based on location, time of day, and groups of individuals, enabling more flexible access management tailored to on-site operations. ▶ Expected Implementation Scenarios - All employees can enter the entrance, while only administrators can enter specific areas at night, etc.

Update of FaceMe Platform 6.4 < New API - Person Attribute Recognition (PAR) > A new API called "Person Attribute Recognition (PAR)" has been added, which can recognize characteristics of individuals based on appearance information other than the face. This allows for the identification of individuals using visual features such as clothing and belongings, even in situations where the face is not visible or difficult to recognize. The following information can be detected: - Age (adult/child) - Gender - Clothing (long sleeve/short sleeve) - Clothing color - Presence of hats or bags < New API - Person Re-Identification (Re-ID) > Another new API, "Person Re-Identification (Re-ID)," can be used to match individuals captured by different cameras and identify the same person. By combining this with the aforementioned Person Attribute Recognition (PAR), it is possible to match the appearance information of a person detected by one camera with footage from another camera, allowing for the identification and tracking of their location and time. This makes it particularly effective for managing wide-area surveillance in places like shopping malls and train stations. Use cases: - Identifying the location of a lost child based on age and clothing information - Tracking suspicious individuals based on gender and belongings information

■Background Yuan Da Bank has adopted CyberLink's online identity verification solution, FaceMe eKYC, in the mobile application process for the "Yuan Da Securities Customer Exclusive Loan Plan" to provide more flexible, convenient, and secure loan services to its securities customers. ■Solution Details This solution integrates technologies for identity document and individual verification (1:1 comparison), authenticity verification of identity documents, fraud prevention, and AI OCR character recognition, enhancing review accuracy and reducing risks associated with forgery and theft, while simplifying and speeding up the application steps. As a result, users no longer need to visit a branch or make phone calls, allowing them to complete the entire application online, enabling loan limits to be established in as little as 15 minutes, significantly improving convenience. ■High-Precision Facial Recognition and Industry-Leading Fraud Prevention Features FaceMe's facial recognition engine achieved an exceptionally high authentication accuracy of 99.83% in the latest FATE test conducted by NIST. Additionally, the fraud prevention feature achieved a 100% fraud detection rate in tests conducted by NIST, earning a top rating and showcasing industry-leading technology.

■Background of Obtaining ISO 27701 Certification CyberLink's AI facial recognition engine, FaceMe, has been continuously strengthening its security management system, in addition to complying with GDPR (General Data Protection Regulation), and aims to obtain the international standard for information security management systems, "ISO/IEC 27001," by 2024. In these efforts, we have aimed to demonstrate a strong commitment to privacy protection and further enhance the reliability of our facial recognition engine by pursuing the acquisition of "ISO/IEC 27701," which establishes requirements for the management and protection of personally identifiable information (PII). ■Proving Commitment to Personal Data Protection with ISO 27701 Certification While ISO 27001 focuses on information security, ISO 27701 emphasizes the protection of users' personal information and compliance with privacy-related laws in various countries. By newly obtaining ISO 27701 certification, it has been recognized by a third-party organization that our company maintains high standards in both privacy protection and information security.

< Update of FaceMe SDK 7.2 > ▶ Performance improvements for anti-spoofing features ✓ Processing speed of 2D anti-spoofing features increased by up to 3 times ・Maintains detection accuracy of anti-spoofing features using RGB cameras on PCs, smartphones, and tablets while increasing processing speed by up to 3 times. ✓ Detection accuracy of IR + RGB anti-spoofing features improved by 67%, with processing speed increased by 2.3 times ・Ensures high detection accuracy even in environments susceptible to external light. ・Target devices: Miyakawa Manufacturing Knoctoi Lite (Noctua Lite), Union UBio-X Pro. ▶ Major update to deepfake detection capabilities ✓ Detection accuracy of deepfakes increased to a maximum of 95.5% ・Adopts a newly trained DNN AI model. ・Sets up two levels of accuracy models: Standard / Fast. ▶ Functionality to detect attributes and clothing from appearance information ✓ Added functionality to detect age, gender, clothing, etc., from appearance information other than the face. ・Even in cases where obtaining facial information is difficult, it is possible to acquire attribute information usable for marketing statistics, etc., from appearance information.

<< FaceMe Security Update >> ■ New Add-on Feature - Unregistered Person Detection ■ ▶ Key Points of the Feature - Detects and issues alerts for individuals not registered in an environment where only registered employees exist. - Contributes to the early detection of potential risks in facilities requiring stringent security, such as factories and data centers. ▶ Features of the Function - Automatically issues alerts when unregistered individuals are detected. - Customizable alert issuance conditions (detection frequency, time intervals, number of days). ■ New Add-on Feature - Intruder Detection ■ ▶ Key Points of the Feature - Works in conjunction with IP cameras, allowing for monitoring area setup, event recording and playback, and alert notifications via API integration. - Offers two functions ideal for safety management in hazardous areas such as construction sites and factories. ▶ Features of the Function - Line Cross Detection: Detects and records events when a person enters a designated area. - Loitering Detection: Detects and records events when a person stays in a designated area for a specified period of time. (Designated areas and loitering times can be easily set on the FaceMe Security UI.)

In this new version, we have improved the UI/UX during VMS integration, enabling smoother system construction. Additionally, a feature has been added to automatically detect ONVIF-compatible IP cameras within the same network, significantly enhancing the efficiency of the initial setup and greatly strengthening technical support. 【 New Features and Updates of FaceMe Security 8.0 】 ＜ Enhanced UI/UX for VMS Integration ＞ - Provides a more intuitive and simple user-friendly UI - Added a guide with clear steps and snapshots ＜ Automatic Detection of ONVIF-Compatible IP Cameras within the Same Network ＞ - Searches and detects cameras within the specified IP address range - Automatically detects all FaceMe components (workstations and add-ons) within the same LAN ＜ Added Trigger Function for External Devices via HTTP API ＞ By customizing the HTTP API on FaceMe Security, direct control is possible for network-enabled devices such as patrol lights and contact control devices for automatic door control.