Custom pre-iBeta testing
Level 2

Case summary: Axon Labs helped Asia’s largest telecom identify inaccuracies in their liveness detection model. Through pre-iBeta testing with Silicon Mask attacks, we uncovered security gaps, enabling the client to address weaknesses early, avoid a failed certification, and strengthen anti-spoofing capabilities.

Task

We were approached by Project manager and Senior AI engineer from leading telecommunications company in South Asia. The client had the task of improving their existing liveness detection model and achieving iBeta Level 2 certifications. Their current model needed significant optimization to meet the strict certification requirements and perform effectively in real-world production environments. It was important for the client to collect a customized dataset: with different lighting, people of different nationalities, genders, etc.

Solution

To ensure the highest level of security for the client’s liveness detection system, we conducted pre-iBeta testing under real-world conditions. Understanding the importance of accuracy and compliance with industry standards, we worked closely with the client, who provided us with access to their application SDK. This allowed us to create a dataset that was close to their production environment while addressing specific business needs.

Key aspects:

  • Silicon mask atacks: We prioritized the most advanced attack type in the Level 2 certification process, ensuring our Axon Labs test effectively simulated the real-world iBeta evaluation
  • Real-world data simulation: We used the client’s SDK to generate a dataset as close as possible to production data, ensuring realistic testing conditions
  • Customized dataset structure: The dataset was adapted to follow a specific structure optimized for this case, allowing for more accurate and meaningful evaluation
  • Risk mitigation: By identifying potential inaccuracies early on, we helped the client strengthen their scam detection capabilities before official certification

This structured approach not only ensured thorough validation of the model but also provided valuable insights to optimize its performance, minimizing risks and enhancing overall security.

Result

  • Stopped a failed iBeta Level 2 certification, saving a lot of time and money. By finding and fixing problems early, the team avoided expensive retries and delays
  • Found weak points in the liveness detection system before the official test. This helped them understand where the system needed improvement
  • Provided high-quality data to help the client prepare for certification. The data made it easier for them to improve their model and create a safer system for users

Accelerate Your AI Development Today ​

Speed up your AI projects with our high-quality, ready-to-use datasets. Enjoy easy integration, fast deployment, and reliable biometric data collection