Online proctoring. How it works? Part II.

Hi there!

In previous part we knew about online proctoring system in general. So now let’s dig into tech staff a bit deeper.

The system includes following modules:

• Workplace compliance inspection module for the proctoring system
• Module for photographing and identification
• Cyber-proctoring module
• Data storing and proctoring session post-verification module
• Control Panel Module

Workplace compliance inspection module for the proctoring system

At the initial stage, the module checks that the camera and microphone are connected and that images and sound can be acquired from both devices.

To make sure that the app can record and transmit video to the server, the module starts broadcasting the video as if it was a real examination. If the module is not able to start the broadcasting process, the student won’t be able to start the examination.

At the same time, a successful broadcasting attempt is not always enough. The quality of the transmitted video may be unacceptable, and the broadcast itself is often interrupted. Therefore, the module checks the quality of the network connection.

Module for photographing and identification

The photography and identification module is responsible for receiving, processing, and storing reference images of the face, as well as images of documents of the user, which have to be stored in the database until the end of the examination session (including the moment of the evaluation of the results of the examination). It can also be stored for any other period defined by the customer.

To pass the stage of automatic identification, the user must take a reference photo, which will later be used by the cyber-identification module during the examination to confirm the identity of the user, as well as for the second and subsequent examinations within the proctoring system.

It’s worth noting that a neural network is used for facial recognition.

Cyber-proctoring module

The cyber-proctoring module consists of three large blocks:

● The face-cyber block is responsible for reading the facial vector and recognizing the direction of view, identifying the face in the frame and the presence of more than one face in the frame.

● Audio-cyber block is responsible for recognizing voice signals in the audio flow of the broadcast.

● The cyber-identification block is responsible to determine if the person participating in the examination is not the same person that was intended to participate.

● The cyber-search block is responsible for searching for a person taking the same examination using different accounts. Such an approach is one of the most popular cheating methods when an excellent student passes the exam five times for himself and four other students.

● The Cyber-desktop block is responsible for controlling the screen of the user and tracking changes to the screen state.

Face-cyber is a neural network that recognizes the face by the face vector: The neural network is trained on a database with thousands of different photos of faces. The neural network recognizes the same face even if one wears glasses, has a beard, long hair or even a different haircut.

Audio-cyber uses a library intended to provide clear recognition of the human voice in a mixed audio stream. When a voice is detected, the system registers the violation and sends it to the control panel of the operator of the proctoring system for evaluation:

Cyber-identification remembers the face of the user who has passed through the photographing stage before the exam, creates the reference photo in the database and periodically compares the face in the frame and the reference photo during the examination. If a person is cheating and pretends to be another person, the violation is automatically registered.

Cyber-desktop uses libraries for frame-by-frame comparison of desktop images. In this case, when there are more than 10% of changes, an automatic violation is detected and sent to the control panel of the operator of the proctoring system for evaluation.

Cyber-search of the same face captures each photo of the face at the stage of photographing, tracks the face vector and searches the database for photo matches on the server-side. In case of any similar photos, information about all of them is added to the report and sent to the customer.

Data storing and proctoring session post-verification module

The data storing and proctoring session post-verification module (archive) contains video files of users (acquired from both the camera and desktop), as well as all violations recorded by the cyber-proctor and operator of the proctoring system.

This module also displays a ‘traffic light’ that evaluates information about the automatic session rating: completed — green, warning — yellow, violation — red.

Violations are also displayed on the video timeline with special colors: the yellow part represents various violations, while the red part represents more than 3 violations per minute.

All information about messages and session violations can be downloaded as a PDF file.

Control Panel Module

This module is available to the customer when working independently with the proctoring system.

In this module, depending on the access rights, the following operations are available:

- Creating exams

- Creating system users: students, proctors, offline proctors, system administrators

- Creating a schedule in a calendar

- View statistics on exams for the day/week

- View the status and manage students during the online examination

- Planning and scheduling operators for online sessions

- Generate reports for all sessions, including statistical reports (number of violators, the most frequent violation, number of participants in the examination, automatic assessment results and so on).

In the following articles, we will talk about various services that are intended to improve the system. We will show you how we usually work with universities and companies, and what you need to take into account when you choose and implement a proctoring system.

Share the most interesting for you topics in the comments.

Don’t forget to subscribe to our blog!

Sincerely yours, Examus Team.