Ijraset Journal For Research in Applied Science and Engineering Technology

Abstract

There had been a giant leap in the field of educations from the past 1–2 years. Education institutes are transitioning online to provide more resources to their students. The COVID-19 pandemic has provided student more opportunities to acquire knowledge and improve themselves at their own pace. These Online proctoring services (part of assessment) are also on the hike, and Artificial Intelligence based Smart Examination proctoring systems (henceforth called as AISEPS) have taken the market by tempest. Online Proctoring system (hence called as OPS), in general, it makes use of internet tools to maintain the sanctity of the examinations. While most of the software uses various modules, the sensitive information they collect raises concerns among the student communities. There are various psychological, cultural and technological parameters that needs to be considered while developing AISEPS. Major issues includes Security, Privacy and ethical concerns, It is difficult to know whether the benefits of these Online Proctoring technology outweigh their risks. Our AISEPS allows users to take exam online in a more secured environment, AI interfaced in the system that allows user to monitor their exam and generate report which is need for hour.

Introduction

I. INTRODUCTION

Over the past few years, online learning has advanced rapidly. More students are taking advantage of these Massive Open Online Courses (MOOCS) and other online certified courses. Colleges are also transforming to online to provide more resources to their students. There has also been a rapid growth in individuals rolling out their courses. All of this offers students more opportunities to study and improve themselves. (Li et al., 2015). In human-based and AI-based systems, we have covered one of a kind sorts that are now in use in the market and the ones suggested in a number of research papers. (O’reilly & Creagh, 2016) Multiple new hardware and software-based improvements have been noted. These are in addition to the number of parameters that are already time-honored as the standard for designing an AISEPS. The pandemic situation has set up an urgent need for a working, AISEPS. This rise in demand also means that the research and development of AISEPS will now be accelerated. These systems use the hardware such as webcams, mics already present in the student's laptop to monitor their activity and ensure academic integrity. Many elements must be taken into consideration while designing a digital proctoring system. The AISEPS must run on all administration without any issues, and it must not be an overly intrusive system. The students are monitored during the exams by faculty through the AI integrated into the system using webcam and mic, with no extra software involved.

II. PROBLEM STATEMENT

A. Objectives

The main objective is to increase the efficient way to conduct exams during pandemic situation which in-turn increases learning. AI based smart examination proctoring system will allows users to fulfil their needs in terms of conducting exam effectively. The flow of the project is as follows. An AI is trained to detect the face and track eyes movement in real-time through live camera feed using tensor flow and YOLO V3(You Only Look Once) model is used for real-time object detection algorithm that identifies specific objects in live feeds, which helps us to detect cell phone and classify as malpractice. This is a web application which allows users to register and then login, During registration user credentials including their present face image is captured and fed into model to train and detect interest points which is then stored into database. Users are loggedIn Only when their face matches with the face captured and stored in database at the time of registration. On Successful login user are directed to their dashboard to perform certain actions. Users are classified as Student and professors. Professor can launch test or assign test to particular student either subject or objective type. Once test is launched an AI model makes sure that Environment is good to take test and monitor the complete test. It generates report at the end of test and stored in database which is accessible by the test provider. The trained model loaded into real time environments.

B. Existing Architectures on AISEPS

ProctorU is an illustration of an OPS that uses a microphone and webcam. It's a live proctoring system in which the proctor attendance scholars through the entire process of an online test, and also monitors them using the webcam. Proctor also needed to corroborate the pupil’s identity by asking the to present their ID cards. scholars are needed to maintain an continued audio-visual connection to the proctor throughout the session. (Milone etal., 2017) Kryterion, a extensively- used marketable OPS uses an approach veritably analogous to the one used by ProctorU (Prathish etal., 2016). The AI module of ProctorU is still isn’t largely secure and can be deceived, which is why the company recommends using their mongrel result to maintain high security. This mongrel result augments automated proctoring with professionally trained live proctors, who have the capability to intrude the test and intermediate in case they suspect commodity.

C. Proposed System

Proposed system is the one which has both backend supported with mysql and front-end with python flask web framework, The entire system runs under apache server. The System allows users to classify based on roles (student or professor) while registering, during registration live image of user stored in database for further process. User (Professor) can lunch exams either subjective, objective or practical a unique test-id will be generated, test-id with password will be shared among test-takers. An AI will detect the face of user through live camera feed using deep face technology using tensor flow. At the end the final report of user test will be available for the user (Professor).

III. METHODOLOGY

A. Front end Technology and Backend Technological Stack Used

• Python Flask web Framework: Python version 3.6.2 is used to build the complete project, the model trained using tensor-flow libraries deep face method to recognize face and matching is done on live. Flask is a web framework, it's a python module that allows you develop web applications easily. It gives more flexibility to developers and is a more accessible framework for beginners since you can build a web application quickly using only a single Python file.
• MySQL Database in xampp Server: As a backend to store the data from users and retrieve the data efficiently we opted for MySql database, which allows us to store images and processed images information using deep face technology. Processed image data, voice data, windows log activity will be stored.

B. Parameters Considered while designing AISEPS

• Camera: This is an input device comes as a part of all laptops as hardware and it is an easily available add-on for desktop systems. The webcam is used as the proctoring authority (PO) with a live view of the user. By this way the user can be monitored to ensure that they attentive in giving their examination, while simultaneously checking for any attempts of cheating. Using deep face recognition technology, the system can make sure that only the registered in user is giving the test. The webcam can also be used to detect more than one person in a room helping to cheat.
• Mic: This is once again an input device that comes with most of the systems. The mic can be used to record audio and process the same. The analysis can then be used to detect whether the user is being assisted by someone out of the field of camera view or via a call on another device. Here background noises can also be considered as dishonest activities, the model needs to be trained well to prevent from false positives.
• Human Proctor: The systems, today we cannot expect to have a hundred percent accuracy rate. These requires human supervision for dealing with false positives and to help with grievance redressal. This way the systems can be improved to better the AI working in the backend. The PO will also analyse the report generated by the AISEPS to render final judgement regarding the malpractice. The AISEPS processes multiple inputs such as Audio, Video and background application data.

C. Python Libraries

Flask_WTF==0.14.3: WTF stands for WT Forms which is intended to give the interactive user interface for the user.

pandas==1.1.5: Pandas is a set of fast, flexible and highly recommended data structures designed to work easily and intuitively with "relevant" or "labeled" data.

tensorflow==2.2.0: TensorFlow is a Python library for fast numerical computing created and released by Google.

stripe==2.27.0: The Stripe Python library which provides convenient access to Stripe API from applications which are written in the Python language.

deepface==0.0.49: Deepface is a lightweight face recognition and facial analysis (age, gender, emotion and race) framework for python.

Werkzeug==0.15.5: Werkzeug has a collection of libraries that can be used to create a WSGI (Web Server Gateway Interface) compatible web application in Python.

Flask_MySQLdb==0.2.0: Flask-MySQLdb provides MySQL connection for Flask application.

Flask==1.1.2: Flask is a web framework, it's a python module that allows you develop web applications easily.

nltk==3.6.2: NLTK is a toolkit build for working with NLP in Python.

object_detection==0.0.3: It is capable of identifying objects that exist in images and videos and tracking them.

Pillow==8.3.2: Pillow is a Imaging Library in python (PIL), which provides support for opening, manipulating, and saving images.

D. Model to Detect Face and Eyes Movements

In this process the model is trained to detect face landmarks by using shape analyser for different faces and different shape, using shape predictor model to detect and track eye movement.

The system is able to track the movement of eyes and detect face once the face appears on the camera. The shape_predictor_68_face_landmarks.dat is integrated to obtain the face landmarks and trace the movement of head and eyes.

E. Object detection YOLO V3

YOLO stands for ‘You Only Look Once’. It is a way of detecting and recognizing various objects in an image (in real time). Object detection in YOLO is done as a regression complex and provides the class probabilities of the detected images.

The YOLO algorithm uses convolutional neural networks (CNN) to detect objects in real time. As the name implies, the algorithm requires only one forward propagation through the neural network to detect objects. This means that the prediction is done in the same algorithm motion throughout the whole picture. CNN is used to simultaneously predict different class probabilities and boundary boxes. The YOLO algorithm consists of various variants.

Here YOLO V3 is implemented to detect mobile phones, when a user try to open his mobile phone in front of camera, the model detects the presence of mobile phone and raises the flags that user is trying to cheat, once the final report about the exam generated mobile phone detection will be marked in red which makes the users to take action on a candidate.

F. Artificial Intelligence Smart Exam Proctoring System

In the process of implementing the system by using the above mentioned sections (A, B, C, D, E) technological stacks, the final system is implemented using html&css integrated with python flask, making use of MySQL database as a backend to store and process user data obtained from the system. The implemented system allows the users to define their role either student or professor while registering with their email-id, email verification can be done by sending a verification code (OTP) to a user email address. During registration user live picture taken as input to store in database for further image verification. Once the registration is successful, user is allowed to login to their destination dashboard. The dashboard is premium and easy to navigate between the tabs.

Professor Dashboard, he/she can lunch exam either subjective, objective or practical and share exam login credentials among students, the created exam is protected by password.

Student dashboard allows the user to start their exam, view results and report any problem that raised while taking exam. AI will monitor the student activities and generate the final report after exam.Once the student finished with his/her exam professor can view their logs activity and their environment condition, based on that professor can insert and publish the marks for student and report any problem that raised 24/7 services will be provided when the website is on live.Stripe payments is used on professor end, as the project is limited only for localhost, when website is on live that allows user to purchase extra 10 exams if he exhausts free exams provided by default for each user. The system implements able to handle the exception that raises while handling errors, built system is efficient and faster.

Conclusion

Based on the results implementation works, the following clarifications can be concluded: AI Based Smart Exam Proctoring System provides: 1) An efficient way to conduct exam online either live monitoring or Automated monitoring. 2) Model can analyse the student environment and generate a report based on decision taken by model as a final report of exam taken. 3) YOLO V3 is able to classify the objects in live feed, raise flag if malpractice is detected, user can launch subject, objective and coding exam for their student in a secure way through online.

References

[1]. Alessio, H. M., Malay, N., Maurer, K., John Bailer, A., & Rubin, B. (2017). Examining the Effect of Proctoring on Online Test Scores. [2]. Arora, P. (2021). Is Remote Proctoring The Future Of Academia? - eLearning Industry. [3]. Atoum, Y., Chen, L., Liu, A. X., Hsu, S. D. H., & Liu, X. (2017). Automated Online Exam Proctoring. IEEE Transactions on Multimedia, 19(7), 1609–1624. [4]. Beust, P., Duchatelle, I., & Cauchard, V. (2018). Exams taken at the student’s home. [5]. Bilen, E., & Matros, A. (2020). Online Cheating Amid COVID-19. [6]. Butler-Henderson, K., & Crawford, J. (2020). A systematic review of online examinations: A pedagogical innovation for scalable authentication and integrity. Computers and Education, 159. [7]. Caveon, D. F., Security, T., & Layman, H. (2013). Online Proctoring Systems Compared. [8]. Chua, S. S., Bondad, J. B., Lumapas, Z. R., & Garcia, J. D. (2019). Online Examination System with Cheating Prevention Using Question Bank Randomization and Tab Locking. [9]. Coghlan, S., Miller, T., & Paterson, J. (2020). Good proctor or “Big Brother”? AI Ethics and Online Exam Supervision Technologies. ArXiv Preprint. [10]. Dendir, S., & Maxwell, R. S. (2020). Cheating in online courses: Evidence from online proctoring. Computers in Human Behavior Reports, 2, 100033. [11]. Draaijer, S., Jefferies, A., & Somers, G. (2018). Online proctoring for remote examination: A state of play in higher education in the EU. Communications in Computer and Information Science, 829, 96–108 [12]. Emerging Technology and the Future of Online Proctoring. (2020). [13]. Furby, L. (2020). Are You Implementing a Remote Proctor Solution This Fall? Recommendations from NLN Testing Services. Nursing Education Perspectives, 41(4), 269–270. [14]. Ghizlane, M., Hicham, B., & Reda, F. H. (2019, December). A New Model of Automatic and Continuous Online Exam Monitoring. Proceedings - 2019 4th International Conference on Systems of Collaboration, Big Data, Internet of Things and Security, SysCoBIoTS 2019. [15]. Golden, J., & Kohlbeck, M. (2020). Addressing cheating when using test bank questions in online Classes. Journal of Accounting Education, 52. [16]. Hussein, M. J., Yusuf, J., Deb, A. S., Fong, L., & Naidu, S. (2020). An Evaluation of Online Proctoring Tools. Open Praxis, 12(4), 509. [17]. Hylton, K., Levy, Y., & Dringus, L. P. (2016). Utilizing webcam-based proctoring to deter misconduct in online exams. Computers and Education, 92–93, 53–63 [18]. Ilgaz, H., & Afacan Adan?r, G. (2020). Providing online exams for online learners: Does it really matter for them? Education and Information Technologies, 25(2), 1255–1269. [19]. Joshy, N., Ganesh Kumar, M., Mukhilan, P., Manoj Prasad, V., & Ramasamy, T. (2018). Multi-factor authentication scheme for online examination. [20]. Karim, N. A., & Shukur, Z. (2016). Using preferences as user identification in the online examination. International Journal on Advanced Science, Engineering and Information Technology, 6(6), 1026–1032.

Copyright

Copyright © 2022 Shalini N, Divakar D, Abhishek H, Bhargav GK, Beemaraya . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.