Enhancing User Authentication and Preventing Fraud with Amazon Rekognition Face Liveness

Learn About Amazon VGT2 Learning Manager Chanci Turner

Permalink

Comments

Share

Various sectors, including financial services, gig economy platforms, telecommunications, healthcare, and social networking, utilize facial verification for online onboarding, enhanced authentication, age-based access restrictions, and bot detection. Users’ identities are confirmed by matching a selfie taken with a device camera to a photo from a government-issued ID or a pre-established profile. Additionally, facial analysis is employed to estimate user age before granting access to age-restricted content. However, bad actors are increasingly executing spoof attacks, utilizing images or videos of users that are publicly available, captured covertly, or synthetically generated, to gain unauthorized account access. To combat this fraud and minimize associated costs, businesses need to incorporate liveness detection into their facial verification workflows before proceeding with face matching or age estimation, ensuring that the individual in front of the camera is indeed a real person.

We are thrilled to present Amazon Rekognition Face Liveness, designed to help you effectively and accurately prevent fraud during facial verification processes. This post will provide an overview of the Face Liveness feature, its various applications, the user experience, an outline of its spoof detection capabilities, and guidance on integrating Face Liveness into your web and mobile applications.

Overview of Face Liveness

Currently, customers use an array of methods to detect liveness. Some employ open-source or commercial facial landmark detection machine learning (ML) models in their applications, requiring users to perform specific gestures like smiling, nodding, or blinking. Such solutions can be expensive to maintain, often failing to defend against sophisticated spoof attacks using 3D masks or injected videos, and demand a high degree of user participation. Others resort to third-party liveness detection features that only recognize spoof attacks presented to the camera, such as printed photos or videos, which may not be effective across all regions and often require significant customer management. Lastly, some solutions depend on hardware-based infrared and other sensors, which are costly and limited to certain high-end devices.

With Face Liveness, you can swiftly verify that users accessing your services are genuine individuals rather than bad actors employing spoofing techniques. Face Liveness boasts several key features:

• Analyzes a brief selfie video in real time to determine whether the user is authentic or a spoof.
• Provides a liveness confidence score ranging from 0–100, indicating the likelihood that the individual is real and alive.
• Delivers a high-quality reference image for downstream face matching or age estimation analysis.
• Offers up to four audit images from the selfie video for maintaining verification records.
• Detects various spoof types, including printed and digital photos, videos, and even advanced deepfake content.
• Easily integrates with applications on most devices equipped with front-facing cameras using pre-built AWS Amplify UI components.

Moreover, no infrastructure management, specialized hardware, or extensive ML knowledge is required. The feature automatically scales based on demand, and you only pay for the liveness checks conducted. Face Liveness employs ML models trained on diverse datasets, ensuring high accuracy across various skin tones, ancestral backgrounds, and devices.

Use Cases

Face Liveness can be applied in several user verification scenarios, as illustrated in the following diagram:

• User Onboarding: Reduce fraudulent account creation by validating new users with Face Liveness before proceeding with further processing. For instance, a financial services provider can utilize Face Liveness to confirm the identity of a user before opening an online account, deterring malicious actors from using social media images to create fake bank accounts.
• Step-up Authentication: Enhance the verification process for high-value actions, such as password changes or money transfers, by employing Face Liveness. For example, a ride-sharing service can use this feature to verify a driver’s identity against an established profile picture, improving safety and discouraging unauthorized individuals from interacting with users.
• User Age Verification: Prevent underage individuals from accessing restricted content online. Online retailers selling tobacco or gambling services can use Face Liveness to confirm a user’s age before granting access to sensitive content, thereby preventing minors from misusing parental information to gain entry.
• Bot Detection: Utilize Face Liveness in place of traditional captcha checks to minimize bot interactions. Social media platforms can leverage this technology to ensure that only real users are engaging with their services, significantly increasing the difficulty for bots to operate effectively.

End-User Experience

When users need to verify their identity on your application, Face Liveness offers a streamlined interface and immediate feedback for capturing a short selfie video. Users are guided to position their face within an oval displayed on their screen. As they comply, colored lights signal successful positioning, and the selfie video is securely transmitted to cloud APIs for real-time analysis. After this analysis, you receive a liveness prediction score, a reference image, and audit images. Depending on whether the confidence score meets the preset thresholds, you can proceed with further verification steps or prompt users to retry or choose an alternate verification method.

The user experience unfolds as follows:

1. The process begins with a start screen detailing instructions and warnings regarding photosensitivity, guiding users to prove their authenticity.
2. Upon selecting “Begin check,” a countdown initiates on the camera screen.
3. Following the countdown, video recording commences, and users are prompted to position their face within the oval. Once correctly positioned, users are instructed to hold still while colored lights flash.
4. The video is submitted for analysis, and a loading screen displays the message “Verifying.”

For additional insights on related topics, check out this blog post here and for compliance guidance, visit here.

For a comprehensive resource on common interview questions, don’t miss this link.