Distance and online learning, often known as e-learning, has emerged as the new norm in training and education in light of recent technological advancements, thanks to a number of benefits like accessibility, cost, efficiency, and usability. Even though the online educational system offers advantages over the traditional educational system, preventing cheating and other improper behavior of students during classes and exams poses substantial difficulties. The \'Artificial Intelligence based Online Examination proctoring system\' project offers a solution to the issues with online and distance education. The system makes use of movement detection, face recognition, and other biometric approaches to flag the student fraud during exams. As soon as the cheating or other irregularities are detected by the students, the system signals the manual proctor and also records these instances as evidence for subsequent monitoring. As a result, online exams can be administered effectively with fewer or no manual proctors present, which is more practical and affordable.
AI-based proctoring systems are technological solutions designed to monitor and supervise online exams or assessments. Traditional in-person proctoring methods are not viable for remote or online education scenarios, so AI proctoring has emerged as an alternative to ensure academic integrity and deter cheating.
AI proctoring systems leverage various technologies such as computer vision, machine learning, and natural language processing to monitor and analyze student behavior during an exam. These systems typically consist of two components: a monitoring component and an analysis component.
The monitoring component uses webcams and microphones to capture audio and video of the students taking the exam. It monitors activities like head movements, eye movements, keystrokes, and background noise to detect any suspicious behavior. Some systems also track facial expressions and gaze direction to identify signs of stress or distraction.
The analysis component applies AI algorithms to the collected data to identify potential cheating instances. It compares the student's actions against predefined patterns and rules, such as detecting if they are looking at unauthorized resources or collaborating with others.
When the system detects a potential violation, it typically generates an alert for human proctors or instructors to review the flagged instances manually. This approach helps minimize false positives and ensures that the final decision is made by a human evaluator.
This 'Artificial Intelligence based Online Examination proctoring system' project offers aim to strike a balance between maintaining exam integrity and respecting student privacy. Overall, it offers a scalable and efficient way to proctor online exams, providing educational institutions with tools to uphold academic honesty in remote learning environments.
The brief methodology of 'Artificial Intelligence based Online Examination proctoring system' involves the following key steps:
Setup and Configuration: The proctoring system is installed and configured on the examination platform or learning management system. This involves integrating the system with the exam interface, setting up the necessary hardware (webcams, microphones), and defining the specific rules and parameters for monitoring.
Exam Monitoring:Once the exam starts, the AI proctoring system continuously monitors the student's activities. It uses computer vision techniques to analyze the video feed from the webcam and detect any suspicious behavior. This includes tracking head movements, eye movements, and gaze direction to identify if the student is looking away or at unauthorized resources. It may also monitor audio for any unusual sounds or background noise.
Data Collection: The system collects data from various sources, such as video footage, audio recordings, keystrokes, and screen activity. This data is processed and stored securely for further analysis.
Behavior Analysis: The collected data is analyzed using AI algorithms and machine learning models. The system compares the student's behavior against predefined patterns and rules to identify potential cheating instances. For example, it may flag instances where the student is accessing external websites, copying and pasting content, or engaging in suspicious eye movements.
Alert Generation: When the system detects a potential violation, it generates an alert or flag for human proctors or instructors to review. This helps minimize false positives and allows human evaluators to make the final judgment based on the flagged instances.
Human Review:The flagged instances are reviewed by human proctors or instructors who make the final decision regarding whether a violation has occurred. They can review the recorded video, audio, and other relevant data to assess the situation accurately.
Reporting and Feedback: The proctoring system generates reports that summarize the exam monitoring results, including any flagged instances and their outcomes. This information is shared with instructors, educational institutions, or relevant stakeholders to ensure transparency and enable appropriate actions to maintain exam integrity.
III. DESIGN AND IMPLEMENTATION
The design and implementation of an AI-based proctoring system involve several key considerations to ensure its effectiveness, accuracy, and fairness. Here is a brief note on the design and implementation aspects:
System Architecture: The well-designed System architecture for the AI based proctoring system is as shown in the Figure 1. It encompasses both the monitoring and analysis components.
Data Collection: The system is capable of collecting data from multiple sources, such as video feeds, audio recordings, screen activity, and keystrokes. This data is captured securely and efficiently to ensure accurate analysis.
Computer Vision Techniques: Computer vision algorithms play a crucial role in analyzing video feeds and identifying student behavior. Techniques like face detection, facial recognition, gaze tracking, and head movement analysis are utilized to detect suspicious activities.
The camera's image input is supplied into the recognition system, where it is efficiently processed and compared. It majorly includes:
a. Image Collection: The suggested system's input is real-time video. The web camera on the user's PC or laptop is used to record the real-time video.
b. Image pre-processing: Pre-processing tries to enhance the quality of the picture data by reducing undesired distortions and boosting some image properties essential for further image processing. Greyscale conversion, noise removal, and picture enhancement are the three main steps in image pre-processing.
c. Image Segmentation: Post image pre-processing, the item is separated from the background image. A black and white image with its contract altered is created to enable better segmentation.
d. Feature Extraction: To extract information from an image, feature extraction is necessary. GLCM is used to capture the spatial dependence between picture pixels. GLCM utilizes the grey level picture matrix to capture the most prominent features. Feature extraction (glcm) tries to suppress the original picture data set by using specific values or features that assist in identifying diverse images from one another.
4. Machine Learning Models: AI-based proctoring systems rely on machine learning models to identify patterns and detect potential cheating instances. These models are trained on labeled data and use techniques like, pattern recognition, and classification to identify suspicious behavior.
5. A Convolutional Neural Network (ConvNet/CNN) is one of the Deep Learning techniques that can take an input image, assign different elements and objects in the image importance (learnable weights and biases), and be able to differentiate between them. Comparatively speaking, a ConvNet requires substantially less pre-processing than other classification techniques. ConvNets are capable of learning filters, but with manual techniques, filters are hand-engineered.
6. Human Review and Intervention:While AI algorithms play a significant role in automated detection, it's essential to involve human proctors or instructors in the review process. Flagged instances are manually reviewed to validate and determine the severity of potential violations. Human intervention ensures fairness, reduces false positives, and allows for contextual understanding.
The suggested \"Artificial Intelligence based Online Examination Proctoring system\" is a user-friendly system built on Python employing the camera/web camera that is widely available with today\'s current computers/laptops, together with CNN, Haar Cascade, and Yolo V3 algorithms. In addition to proctoring, the system is dependable and offers students taking the exam a great user experience. Students who exhibit questionable behavior are flagged by the system. Additionally, it provides real-time feedback to instructors and students. The project also keeps track of any logs of questionable behavior for future reference. By efficiently overseeing the online exam and significantly lowering the amount of human labor required for proctoring, the suggested method will tremendously assist in the delivery of online education.
 Motwani, Sahil, Chirag Nagpal, Manav Motwani, Nikhil Nagdev, and Anjali Yeole. \"AI-based proctoring system for online tests.\" In Proceedings of the 4th International Conference on Advances in Science & Technology (ICAST2021). 2021.
 Sridhar, Arjun, and J. S. Rajshekhar. \"AI-integrated Proctoring system for online exams.\" Journal of Artificial Intelligence and Capsule Networks 4, no. 2, pp. 139-148, 2022.
 Khanna, Vishesh, Sahil Brodiya, and Deepesh Chaudhary. \"Artificial intelligence based automated exam proctoring system.\" Int. Res. J. Eng. Technol. (IRJET) 8, no. 12, pp. 558-560, 2021.
 Guney, Caner, Onur Akinci, and Kadir Çamo?lu. \"Artificial learning-based proctoring solution for remote online assessments:\" vproctor\".\"ISPRS Annals of Photogrammetry, Remote Sensing & Spatial Information Sciences, 2021.
 Masud, Mohammad M., Kadhim Hayawi, Sujith Samuel Mathew, Temesgen Michael, and Mai El Barachi. \"Smart online exam proctoring assist for cheating detection.\" In Advanced Data Mining and Applications: 17th International Conference, ADMA 2021, Sydney, NSW, Australia, February 2–4, 2022, Proceedings, Part I, pp. 118-132. Cham: Springer International Publishing, 2022.
 Yelkar Anjali Rajendra, Pawar Reena Vishwas, Khopkar Diksha Anil, Ratwadkar Pallavi Pravin. “Online exam proctoring system”. International Journal of Creative Research Thoughts, vol 10, Issue 1 January 2022.
 Mohan Vamsi A, Niteesh B, Sai Ashwin D, K Kajendran. “REMOTE ONLINE PROCTORING SYSTEM “. International Journal of Creative Research Thoughts vol 9, Issue 5 May 2021.
 Sinha, Prakash, and Aman Yadav. \"Remote proctored theory and objective online examination.\" International Journal of Advanced Networking and Applications 11, no. 6, pp. 4494-4500, 2020.
 Ahmad, Istiak, Fahad AlQurashi, Ehab Abozinadah, and Rashid Mehmood. \"A novel deep learning-based online proctoring system using face recognition, eye blinking, and object detection techniques.\" International Journal of Advanced Computer Science and Applications 12, no. 10, 2021.
 Dilini, Nimesha, Asara Senaratne, Tharindu Yasarathna, Nalin Warnajith, and Leelanga Seneviratne. \"Cheating detection in browser-based online exams through eye gaze tracking.\" In 2021 6th International Conference on Information Technology Research (ICITR), pp. 1-8. IEEE, 2021.