With the rise in sophistication of forgery attacks, it became necessary to include techniques that verify whether a presented face of an individual is real or a digital recreation. As such, the fusion of facial biometrics and proof of life techniques emerges. The liveness detection technology has two main methods: the passive and the active.
In this article, we will show the difference between these types of liveness technology and how they can contribute to identity security.
Passive or active Liveness Detection
There are differences between active and passive liveness. Let’s see them below.
Passive Liveness
To determine the authenticity of biometrics, passive liveness (or passive facial liveliness) relies on analyzing intrinsic characteristics within facial images such as texture, movement, and depth.
In this method, Machine Learning algorithms are applied to identify patterns that separates real images from fake ones. This approach involves analyzing subtle movements and facial features that indicate the image belongs to a living person, rather than a pre-recorded photo or video.
Passive liveness adds an additional layer of security to facial authentication without requiring the user to interact beyond what is needed, for instance, it can be used in public places like airports, train stations, and subways. As a result, this method becomes quicker for the user.
Active Liveness
While passive liveness does not require direct user interaction, active liveness demands user cooperation for authentication.
The technique involves requesting specific user actions, such as blinking, smiling, or moving the head. Captured by sensors such as infrared cameras or depth sensors, these actions are analyzed to verify facial authenticity.
This method is considered more robust in terms of security,since it is more challenging to bypass.
In summary, passive liveness is convenient and quick, while active liveness is more secure and resilient.
Most common liveness detection applications
Active liveness requires specific user actions, making it more suitable for critical security applications with in-person requirements, for instance:
- Access control in restricted areas.
- In in-person authentication for financial transactions.
However passive liveness allows for a smoother user experience, since the customer doesn't need to perform extra actions. Thus, it's widely used in applications where convenience and speed are crucial, such as:
- Unlocking smartphones.
- Authenticating payments in mobile applications.
- Preventing fraud in identification systems.
- Controlling access to highly secure virtual locations.
How to implement Passive Liveness
The implementation of passive liveness involves computer vision algorithms and machine learning for analysis and proof of life identification. While the exact implementation varies depending on the specific technologies and algorithms used, we can list the main steps involved:
- Data collection: Feeds the system with examples of real and fake faces. This includes images and videos of real people, as well as static photos or videos. It's important to have a variety of examples to represent different scenarios and conditions.
- Data preprocessing: Prepares the data for analysis with operations like image resizing, color normalization, and noise removal.
- Feature extraction: Uses computer vision algorithms like image descriptors or convolutional neural networks (CNNs) to extract relevant features from the faces in the collected data examples.
These features may include information about facial textures, subtle movements, blink patterns, and other distinctive elements.
- Model training: Uses prepared datasets and extracted features to train a machine learning model, such as a binary classifier. The goal is to teach the model to distinguish between real and fake faces based on the identified features.
- Model validation and tuning: Evaluates the accuracy and performance of the model with a separate validation dataset. Adjustments to model parameters and feature selection are necessary to improve performance, if needed.
- Testing and continuous evaluation: Performs rigorous testing of the trained model on a separate test dataset to verify the effectiveness of the liveness detection system.
Monitor the model performance in different scenarios and conditions and make continuous improvements if necessary.
- Integration: Integrates the liveness model into an application or system. This might involve implementing specific APIs or libraries to perform real-time analysis of faces captured by the device's camera.
With all the steps mentioned above, Vsoft is able to deliver its best technology against fraud using Liveness Detection in the BioPass ID product. As a result, the security of data and biometric identifications will be assured that there will be no fraud in the systems.
Now, it’s your turn
Before performing the Liveness Detection, it's important to emphasize that the implementation may be a technical challenge, and there are several available approaches.
Therefore, it's advisable to seek academic literature, scientific articles, and code examples to comprehend best practices and the most suitable technologies for specific applications.
Would you like to explore a practical example of how the liveness detection technology works? Read about how liveness is employed by the Brazilian National Institute of Social Security (INSS) to prevent fraud.
Tradução: Pedro Garrafielo
