Playground biometrics demo BioID home page

Classic BWS - Release Notes

Thursday, May 2, 2024

  • Liveness detection update + latest Deepfake detection
  • Changed handling of EXIF Rotation Tag in images

Wednesday, December 15, 2023

  • Liveness detection now includes popular Deepfakes

Thursday, January 12, 2023

  • Improved liveness detection, especially for mask attacks (silicon, latex, resin, etc.).

Tuesday, September 13, 2022

  • Improved liveness detection (especially for silicon masks)
  • Better PhotoVerify performance with high-resolution images

Tuesday, June 14, 2022

  • Improved liveness detection (especially for detecting masks)

Wednesday, March 23, 2022

  • Improved liveness detection

Tuesday, February 8, 2022

Verify 4.0
  • New face verification/identification algorithm
  • Enhanced overall recognition performance
  • Now, Verify API uses same neural networks as PhotoVerify

Monday, January 24, 2022

This version includes the following improvements:
  • Liveness detection improvements for optimized BPCER/FRR and APCER/FAR:
    • Revised motion field analysis resulting in higher tolerance towards accidental device movement
    • Enhanced live and fake datasets

Wednesday, November 24, 2021

This version includes the following improvements:
  • Improvement of PhotoVerify. Therefore, better FAR rates are implemented now:
    • Level 5: 0.001% FAR
    • Level 4: 0.01%  FAR
    • Level 3: 0.1%   FAR
    • Level 2: 0.25%  FAR
    • Level 1: 0.5%   FAR

Monday, October 4, 2021

  • Liveness detection improvements

Thursday, September 2, 2021

  • Liveness detection improvements
  • New liveness detection sample errors: in case a liveness detection fails (i.e. the operation returns one of the error codes LiveDetectionFailed or ChallengeResponseFailed), the samples may contain new additional error codes as follows:
    • UnsuitableImage: The face is not fully visible in the image.
    • UnnaturalMotionDetected: The 3D detector observed some unnatural motion.

Tuesday, June 1, 2021

  • Face finder improvements
  • Liveness detection improvements

Monday, April 19, 2021

  • Liveness detection improvement for certain desktop cameras

Thursday, January 28, 2021

Verify 3.0
  • New face verification/identification algorithm
  • Less enrollment samples required
  • Enhanced overall recognition performance
  • Now, Verify API uses same neural networks as PhotoVerify

Monday, January 18, 2021

  • Liveness detection improvement for low-res cameras

Tuesday, January 12, 2021

This version includes the following improvements:
  • Improvement on liveness detection
    • Higher robustness concerning medical masks
    • Extended fraud and bona-fide datasets
    • Overall enhanced liveness detection performance
  • Less enrollment samples required
  • Enhanced overall recognition performance
  • Now, Verify API uses same neural networks as PhotoVerify

Thursday, October 22, 2020

This version includes the following improvements:
  • Improvement of liveness detection
  • Improvement of PhotoVerify. Therefore, better FAR rates are implemented now
    • Level 5: 0.01% FAR
    • Level 4: 0.1%  FAR
    • Level 3: 0.25% FAR
    • Level 2: 0.5%  FAR
    • Level 1: 1.0%  FAR

Wednesday, August 19, 2020

This version includes the following improvements:
  • Expansion of fraud prevention scenarios
  • Improvement on mask rejection for various materials
  • Finetuning of liveness detection to lower false rejectionss
  • The RESTful Wep APIs PhotoVerify and LiveDetection use extended return objects, which now also contain quality check and liveness detection results for the provided live images.

Monday, June 22, 2020

This version introduces a completely revised face finder. Cluttered background or asymmetric lighting is a problem of the past now.

Therefore, the error message “no face found” should really only appear if there is no visible face within the picture.

This affects all of our biometric services – the false rejection rate is lowered by this improvement.

Monday, April 6, 2020

This version implements a major improvement of our algorithms for face and periocular recognition, internally called "Verify 2.0":

  • New biometric feature extraction for both face and periocular traits
  • The new algorithms yield much better recognition accuracy, bringing down both FAR and FRR
  • Both Verify and Identify will automatically make use of the new features (existing ones have been ported automatically)
  • Using Verify, which uses a fixed threshold for its decision, with the new algorithms there will be much less false rejections from now on

Since version 2.9.20105.1 the response of all RESTful Web API calls contains an additional response header describing the installed BWS-Version.

Tuesday, January 14, 2020

This version includes the following improvements:

  • Liveness detection: Expansion of fraud prevention scenarios
  • Improvement on silicon mask rejection for liveness detection
  • Finetuning of PhotoVerify accuracy levels and reduction of false rejections
  • Enhanced framework performance

We changed our BWS platform to: .NET Framework 4.7.2, running on Windows Server 2019, therefore please note that TLS 1.2 is now required!

Wednesday, July 31, 2019

Due to requests from our customers we added two new APIs and we added an optional parameter to some Web APIs:

  • PhotoVerify2 is a new version of PhotoVerify, which has the same arguments but returns an object that also contains the calculated actual level of accuracy.
  • VideoLiveDetection is a new face liveness detection API that receives a video as input. The algorithm is identical to the face liveness detection of single images, but as we can choose from more images, VideoLiveDetetcion should be a bit more robust than LiveDetection on only two images.
  • PhotoVerify and VideoLivenessDetection have also been implemented with the SOAP interface. Just update your service reference to get access the new BWS service contract.
  • The Web APIs QualityCheck, PhotoVerify, PhotoVerify2, LiveDetection and VideoLiveDetection have and additional, optional parameter state, which, if provided by a client, is simply passed through to the BWS log and to the returned object, i.e. it is just a piece of additional information typically used for logging purposes.

Wednesday, January 9, 2019

There has been a major upgrade for Face Liveness Detection. With newly optimized DCNNs (Deep Convolutional Neural Networks) we now have a more than 20% higher accuracy in distinguishing live from fake faces. And it even got faster. There are no changes in using the API – no code modifications are necessary for you to benefit from our newest technology.

Thursday, August 9, 2018

Face Liveness Detection has been enhanced significantly. Until now, the liveness detection used 2 images to calculate the 3D shape of a face (and compared this to a natural face geometry). Starting with version 2.4 of BWS, we added a "texture based" algorithm, which is trained to distinguish real human skin from artificial surfaces like a display. With that new development, the BWS Liveness Detection can reject fakes coming from photos, videos, or avatars.

Please note that you need to supply colour images now to use the texture based enhancement (if only greyscale images are sent, only the 3D geometry algorithm can be used). Have a look at our Github examples to see which changes are necessary in the code to make full use of our new Liveness Detection.

Tuesday, June 6, 2017

Although we had some minor updates and bug-fixes in the last two years, we increased the version number only now, as we added some new features:

  • A new trait has been introduced: periocular. The periocular trait (referred to in marketing literature simply as "eye recognition") differs from the face trait in that it looks only at the fine features around around the eye, such as lashes, lid, brow, and folds in the skin, as well as full or partial iris (the complex patterns on the colored part of the eye).
  • Two new useful Web APIs: a simple face liveness detection on two images and a more enhanced photo verification, which performs a liveness detection on uploaded samples and then makes a one-to-one comparison with a ID photo in order to verify whether the live images and ID photo belong to the same person.

Wednesday, May 6, 2015

We released a new version of the BioID Web Service (BWS) and the BioID Playground demo system.

The major changes in this release are:

  • Fix of a REST basic authentication bug when authentication header was missing (we accidentally captured the 401 - Unauthorized response on the server side and returned 200 - OK)
  • Fix of some issues in the REST API when used with the voice trait - the RESTful Web API now supports WAVE (RIFF-based waveform audio file format) data, PCM encoded, 22050 Hz, with 8 or 16 bits per sample, mono or stereo
  • Update to Azure SDK 2.6 (major changes in diagnostics)
  • We added some additional SOAP diagnostics capabilities
  • Due to requests by clients, we added an isEnrolled method to the SOAP and REST API
  • Some algorithm changes we won't disclose here ;-)

Friday, October 10, 2014

We released a new version of the BioID Web Service (BWS) and the BioID Playground demo system.

With this version we restructured our service implementation:

We now use a single Azure worker role (something like a Windows Service) to host all our BWS services, i.e. we self-host a WCF service host for the SOAP version of our BioID Web Service and we self-host an OWIN (Open Web Interface for .NET, see host to run the RESTful BWS Extension. This makes our live a bit easier and the services a bit faster.

The major changes in this release are:

  • All services are self-hosted in a Windows service (worker role).
  • No IIS is used any anymore! This leads to a faster response time and no wake-ups are needed.
  • The unified user interface is not longer part of the BWS Extension. See the BWS Unified User Interface documentation for more information..
  • As usual our algorithms are getting better and faster :-)

Tuesday, December 3, 2013

A new version of the BioID Web Service (BWS) and the BioID Playground demo system was released!

This version has several major internal restructurings:


The major changes in this release are:

  • Enhanced information about possible recognition problems
  • Faster processing because face finding only needs to be done once
  • No more inconsistencies with LiveDetection (which used a different face finding)


  • Very small movements are now also accepted
  • Higher recognition rate
  • Faster processing
  • Explicit error message when two identical images are given


  • Image samples can now be tagged with a forced movement direction
  • BWS methods enroll/verify/identify use those tags and return ChallengeResponseFailed if the directions are not correct


  • Service now reports the reason if a wave file is not accepted (e.g. sampling frequency too low)


  • Rotation parameters were slightly imprecise in former versions, leading to a few pixels inaccuracy
  • The fix affects BioID apps ExtractFace and Pic4Pass
  • Processing mode ICAO adapted to yield pixel-precise results

Thursday, August 29, 2013

A new version of the BioID Web Service (BWS) and the BioID Playground demo system was released! The new version 1.8 fixes a lot of bugs and also a lot of improvements for performance and accuracy have been incorporated in this version.

But the most notable changes are

BioID Web Service

  • We achieved vast performance improvements of the identify operation.
  • A new auto-enrollment flag is now introduced for the verify operation.
  • With the new BWS Extension a RESTful web API is available for the BioID Web Service (BWS)
  • For API authentication the BWS Token using a JSON Web Token (JWT) mechanism is used.

Thursday, March 28, 2013

We just deployed a new version 1.7 of the BioID Web Service and BioID Playground demo system!

The most obvious changes and improvements are

BioID Web Service

  • The voice trait is now using a noise suppression filter which enhances noisy signals, leading to less "Bad Quality" rejections.
  • The pattern management at the BioID Playground profile page is officially released now ( see post Manage your Patterns on the BWS Blog).
  • The credentials page in the BWS Portal for an BWS instance allows to register and manage BWS client applications.
  • Specification, implementation and documentation of a RESTful API representing the BWS Token Service.
  • The sign-up form for new trial instances of BWS is now hosted on the BioID Playground also.
  • The EnrollmentFlags.OverwriteExisitingData is now officially obsolete! See the API Documentation for more information.

Thursday, November 29, 2012

We just deployed a new version of the BioID Web Service and BioID Playground demo system. This is a minor release so the BWS version remains at 1.6.

Most noteworthy changes and improvements are

BioID Web Service

  • New compact layout for BWS management.
  • Accounting view now with audio and image log sample viewer.
  • Performance monitor for CPU and memory.
  • Azure SDK updated to version 1.8 on development / build systems.
  • Arbitrary API support on BWS web roles implemented.

Playground demo server

  • Simpler BWS trial certificate and instance creation process.
  • BWS Blog finally released for public access.
  • Audio sample amplitude graph display in Silverlight audio capture control.

Monday, September 10, 2012

We just deployed a new version of the BioID Web Service demo system into production. As usual it can be accessed via the URL

This is not a whole new release so the BWS version is still at 1.5.

Most noteworthy changes and improvements are

BioID Web Service

  • Enhancements of the BioID Web Service Management Portal
  • Re-submission problem of the QualityCheck method was fixed

Playground demo server

  • Scheduled off-site backup of logging / performance data implemented
  • Enhanced and restructured management controls for BWS administrators.

Thursday, August 9, 2012

Version 1.5 contains a new implementation of the internally used classifiers which increase the performance of enrollments, verifications and especially identifications. We now should be able to support up to 50.000 users with a single identification instance.

The face-detection algorithm again has been improved (although it might be a bit slower now, we can find more faces with the same distinguished accuracy as usual).

In addition to some minor non-breaking changes, we also made a little breaking change by adding the live detection capabilities to the enrollment process. This led to an additional enrollment-flag, which has been added to the schema without changing the namespace (in case you are using existing code generated from the WSDL below, please re-create this code if you want to use the new LiveFaceDetection flag).

Thursday, August 9, 2012

With version 1.4 we implemented some performance improvements and a few new features. Some minor bugs have been fixed as well. Anyway, there are no breaking changes in the interface or schemas.

This version again improved the photo fake defender, which is now more reliable regarding live face detection (without a loss of accuracy in photo fake detection). As a new feature, the live detection now calculates the head movement direction reporting the degree of head rotation between two subsequent images. The photo fake defender problems for images in panel format, e.g. from an iPhone, have been fixed.

We also improved the identification process regarding speed and memory management. So we can now support up to 20.000 (and increasing) users with a single identification instance. As the update interval for processing changes of the enrolled users is now configurable, the identification service can better be tailored to the needs of the customers.

Wednesday, February 29, 2012

Today we released the first version of the BioID Web Service (BWS), that also contains the identification API! We also added some major improvements regarding the photo fake detection.

Beside of making the identification API available (see BWS Identification API), we added some major improvements regarding photo fake detection (live face detection), which can be activated for verification and for identification.

We also modified the signature of the QualityCheck method but without changing the namespace, as this is the first official implementation of the quality check method (see BWS QualityCheck API).

But anyway, we made some breaking changes to the message schemas, especially for the return messages to make it easier to handle and to extend these messages. Refer to the schema for BWS messages for a description and some sample code.

Tuesday, November 15, 2011

As of today, November 15, 2011, we released the first official version of the BioID Web Service (BWS), a cloud-based online service providing a multimodal (face and voice) biometric recognition technology.

Although not all methods of the BioID web service interface are fully implemented yet, we officially released the current version 1.2 of the BWS. Please note that this version has some breaking changes! Therefore it uses modified namespaces, i.e. developers that used a pre-release version of BWS need to update their BWS code.

Implemented Features

  • Enrollment support for face and voice trait
  • Verification support for face and voice trait
  • Request status information containing running instances and enabled features
  • Delete a class from the storage