Although AI has become a powerful tool helping companies gain a competitive advantage, advances in AI technology are also making AI-generated attacks such as Deepfake rapidly mature, posing significant threats to identity verification and cybersecurity. According to a research report, Deepfake incidents surged tenfold from 2022 to 2023. The world today faces an unprecedented cybersecurity challenge. For example, in a Deepfake scam, a finance worker at an international corporation's Hong Kong office was tricked into transferring US$25 million to cybercriminals.
According to Yvonne Lin, Head of Marketing at Authme, there are in general two types of Deepfake attacks. With the first type, the fraudster sets up a mobile phone or PC in front of a webcam and displays a Deepfake image or video. The person on the other side of the video call will unlikely be able to tell the difference. The second type is video replacement. The fraudster hacks into the intermediate network layer and intercepts and replaces the image or video in transmission so the person on the other side sees a Deepfake. More and more developers and researchers are making efforts to counteract Deepfake attacks. Authme's face recognition techniques enable real-time detection of Deepfake-generated images, safeguarding companies across various sectors against Deepfake scams.
AI liveness detection achieves 99.7% accuracy
Two of Authme's founders were once white hat hackers. They also started a cryptocurrency exchange together. From their experience, they found all sorts of pain points with traditional identity verification techniques. This drove them to establish Authme in 2019 with an aim to address the challenges in digital identity verification for companies. During the early-stage development of Authme, the founders participated in a great number of digital identity policy discussions. They also worked with former Minister of Digital Affairs Audrey Tang to fix vulnerabilities in government platforms. For the team's enthusiasm and industry-leading technology, Authme has been recognized and trusted by government agencies and financial institutions over the past three years.
Lin noted that Authme's face recognition includes AI-based face comparison and liveness detection. All it needs is a simple selfie to complete identity verification, effectively preventing identity theft and enhancing account security. Authme's face comparison is NIST FRVT 1:1 certified, achieving a 99.7% recognition rate in the VISA-BORDER scenario. Its passive liveness detection is ISO-30107 certified. Utilizing AI liveness detection to analyze biometric features including skin texture, facial depth, and blood flow, Authme's solution achieves a 99.7% accuracy in detecting whether it's a real person in front of the camera or not, thereby spotting a Deepfake video or image shown in front of a webcam.
Authme was founded to support the finance sector that sets a high standard for and attaches great importance to identity verification. For its unparalleled technologies, Authme's solutions have been adopted by multiple financial institutions. On the basis of a great number of successful use cases in the finance sector, Authme's solutions have further been adopted by cryptocurrency exchanges and vehicle rental companies.
To improve its customer service quality, Authme uses Amazon Elastic Compute Cloud (Amazon EC2) from Amazon Web Services (AWS) to boost the performance of its face recognition solution. On top of that, it adopts the AWS Cloud Hardware Security Module (AWS CloudHSM) for data encryption. Moreover, AWS Identity and Access Management (IAM) is utilized to securely control access to AWS resources, ensuring that only authorized users and services have the minimum required permissions to access specific resources. This helps minimize potential risks if accounts are compromised. Furthermore, Authme leverages AWS CloudTrail, a service that records API calls and related events, to log every activity that occurs in the system, allowing for real-time tracking and instant detection of any suspicious access attempts, thereby effectively controlling risks.
Leveraging generative AI, Authme forays into Asian markets
Authme founders grew up in Kaohsiung but Authme's operation has largely focused on northern Taiwan since its establishment. When they learned of the Startup Terrace Kaohsiung AWS Joint Innovation Center (JIC) program, they immediately decided to join in hopes of strengthening connections with companies based in southern Taiwan and tapping into more business opportunities through the project. Moreover, Authme is also actively expanding into Japan, Indonesia, and other countries in Asia. As part of its overseas expansion efforts, Authme looks to make its products available through AWS Marketplace, a curated digital catalog that makes it easy for customers to find, buy, deploy, and manage the third-party software they need to build solutions and run their business.
According to Lin, aside from its AI face recognition solution, Authme also offers an AI OCR ID document verification solution that supports the recognition of text in multiple languages, punctuation marks, and common mathematical symbols within just 1 second. Trained on a dataset of over 30,000 Chinese characters, it achieves a remarkable 99% accuracy in recognizing Chinese text. Authme is endeavoring to incorporate generative AI into its OCR ID document verification solution to increase the accuracy rate for text recognition of various languages as a way to boost its international competitiveness.
It is noteworthy that as remote work has become the norm, global cybersecurity incidents such as identity theft have been on the rise. In view of this, in 2024, Authme launched its Zero Trust solution, including single sign-on (SSO) and multifactor authentication (MFA) schemes. It is a secure and high-performance identity verification solution for companies, allowing employees to securely and seamlessly access corporate resources and focus attention on tasks that are critical to business operations.
Authme's AI liveness detection accurately identifies the face in front of the camera as being fake in a simulated Deepfake attack
Credit: Authme