Automated Proctoring System Using Machine Learning for Remote Exam Integrity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing use of electronic devices during examinations poses challenges for proctors to detect cheating, especially in remote examination settings where real-time monitoring is difficult.
Innovation Solution
An automated proctoring system that uses image and video analysis, combined with machine learning models, to verify the identity of examinees and monitor the examination environment for prohibited and required items, allowing for remote proctoring and reducing the need for physical presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If remote proctoring is implemented to allow examinations at home, then examinee convenience and accessibility are improved, but the ability to detect cheating deteriorates
Solution Approach 1:
The system performs preliminary actions by capturing images and videos of the examination environment and examinee before the examination begins. Environmental scans detect prohibited items, and body scans verify the examinee is not hiding devices. This preliminary verification establishes a baseline for detecting cheating during the examination without requiring continuous intrusive monitoring.
Solution Approach 2:
The system creates visual copies of the examination environment and examinee through camera feeds and image capture. These copies are transmitted to proctors for remote review, allowing proctors to visually inspect the environment and examinee behavior without being physically present. This copying approach enables remote proctoring while maintaining detection capabilities.
2Reliability
If proctors manually search examinees for prohibited items, then cheating prevention is improved, but time consumption and efficiency deteriorate
Solution Approach 1:
The system replaces the manual mechanical search process with automated image recognition technology. Machine learning models analyze images and videos to automatically detect prohibited items such as smartphones, smartwatches, and earpieces. This substitution maintains high reliability in cheating prevention while dramatically reducing the time required compared to manual searching by proctors.
3Measurement precision
If comprehensive environmental scanning is performed to detect prohibited items, then cheating detection accuracy is improved, but system complexity and processing requirements deteriorate
Solution Approach 1:
The system segments the prohibited item detection task into specialized machine learning models for different types of devices. Separate models are trained to detect smartphones, smartwatches, earpieces, and other prohibited items. This segmentation allows each model to focus on specific characteristics of particular device types, improving detection accuracy while managing system complexity through modular architecture.
Data Source
AI summary
A system is provided for performing a validation of an examination environment. The system acquires a video of the examination environment. The system applies one or more machine learning models to images (frames) of the video to indicate whether the image includes a prohibited item. A machine learning model may be trained using images of items labeled with an indication of whether an image includes a prohibited item. The system determines whether the validation has passed based on whether an image includes a prohibited item. The system notifies a proctor of when the validation has not passed and provides to the proctor an indication of an image that contains a prohibited item. The proctor then decides whether the validation should pass or fail.