In this release, we’ve implemented the following features:
We’ve added functionality to verify the test-taker’s identity using a reference photo from the LMS. This preliminary check runs automatically in the background and does not require any additional action from the test-taker.
If the system cannot confirm the identity, it sends the corresponding status to the session but does not prevent the exam from starting.
All warnings confirmed or manually added by a proctor during the session are now visible in the Sessions Report.
A new parameter, Violations confirmed by Proctor, is available on the report generation page.
The learner web application has been fully translated into 10 additional languages:
We’ve upgraded our device detection approach from classification to object detection to enhance its accuracy and performance.
The new model detects 87% of objects across all frames (based on the mAP50 metric) and operates 40% faster than the previous version.
This updated approach enables more reliable detection across different device classes and allows continuous tuning of model parameters based on customer feedback and new data.
We’ve improved the model used to detect violations in side-view videos (captured by auxiliary or side-view cameras).
The updated model reduces false-positive alerts for detected devices and improves the accuracy of head detection.
We’ve expanded the list of applications that are automatically blocked when a test-taker attempts to run them while using the Secure Browser in Medium and High modes.
The update includes a broader range of screen recording and similar applications to enhance exam security.
We've fixed the following known issues: