TECHNICAL REPORT
| Grantee |
Universiti Kebangsaan Malaysia
|
| Project Title | Combating Cyberbullying for Better Internet Inclusivity: An Optimized Deep Learning Approach |
| Amount Awarded | USD 30,000 |
| Dates covered by this report: | 2023-01-01 to 2023-12-30 |
| Report submission date | 2024-04-24 |
| Economies where project was implemented | Malaysia |
| Project leader name |
Mohd Asyraf Zulkifley
|
| Project Team |
Esther Siok Yee Tan
Fadzilah Othman
Rushda Muharar
|
Project Summary
Cyberbullying is a threat to diversity on the Internet, because certain classes of people can be denied safe access - this may be because of race, religion or gender.
This project identified three key objectives to combat cyberbullying:
- Conduct 13 awareness campaigns (the first 13 out of 17 sessions) on the cyberbullying phenomenon to diverse target groups
- Develop an intelligent cyberbullying detection tool based on an optimized deep learning network
- Integrate the cyberbullying detection tool into the promotional campaign (the last 4 out of 17 sessions)
For the first objective, the project conducted 17 awareness sessions that targeted primary school students, secondary school students, and university students. It also completed our development of an AI-based cyberbullying detection based on text input.
From the findings, the best model was a one-layer CNN coupled with two fully connected layers with 82.26% accuracy. The input data is based on text data of various cyberbullying incidents, which was saved in the CSV format taken from Twitter, where the dataset is then preprocessed before they are tokenized. The data is then split into 75% of training data and 25% of testing data which were randomized at each run of training the model. The tokenized data is then fed into the deep neural network model where the training process is executed until the model loss converges to zero. The optimal base architecture is first determined by analyzing several different architectures, especially the number of layers and the building block of the model.
The optimal hyperparameters are determined by testing individual hyperparameters in sequence by assuming a greedy search approach, which is optimized separately and carried forward for each subsequent test. Finally, an attention mechanism is embedded to enable the model to focus on key phrases in the input sequence to increase the model’s classification accuracy. This developed system and activities have been explained in TWO manuscripts, which are currently under review at Heliyon Journal with the title “Cyberbullying Detection using Attention-based Deep Neural Networks” and recently accepted "Campaign on Awareness of the Negative Impacts of Cyberbullying''.
Besides that, the project also organized four workshops (the last 4 out of 17 sessions) to raise awareness of cyberbullying through technological tools implementation that include advanced artificial intelligence methods to identify cyberbullying incidents. Two primary and two secondary schools were involved in this active activity that include team sessions so that the students really understood how to identify and overcome cyberbullying. One important was that is if the participants are part of Peer Mentoring Prefects, the impact of the project can be larger, whereby they have acted as the motivator to their friends in combating the cyberbullying issue. Furthermore, the project developed a dedicated website to share our findings that cover various types of material for public access.
Table of Contents
- Background and Justification
- Project Implementation Narrative
- Project Indicators
- Project Review and Assessment
- Diversity and Inclusion
- Project Communication
- Project Sustainability
- Project Management
- Project Recommendations and Use of Findings
- Bibliography
Background and Justification
One of the initiatives championed by the Universiti Kebangsaan Malaysia is holistic and inclusive education for all. With the rebranding of UKMShape which focuses on online massive learning, a lot of new interesting behavioral patterns emerged, especially during the COVID-19 pandemic period. Some of the patterns can lead to negative consequences or to a certain extreme, cyberbullying. In some reports, students feel that they are being alienated due to various factors. Coincidently, the same problem was shared by various parents during the online learning session at home. Hence, the awareness of cyberbullying is considerably low among Malaysians, whereby a lot of negative interactions are considered as just normal disagreements. In fact, some of them are even advised to toughen up, instead of acknowledging the occurrence of bullying.
In line with the Industrial Revolution 4.0, the university has encouraged the development of intelligent tools to help reduce the incidence of cyberbullying through deep learning technology. As a result, an AI-based tool enables many parties to recognize and detect cyberbullying incidents effectively, which can also act as a monitoring tool to prevent this problem from escalating to a more severe consequence such as suicide.
To be precise, two main problem statements are used to summarize the motivation behind the proposed project:
Problem Statement 1: Internet users, especially the younger generation are not aware of cyberbullying, in which the aggression is not done physically. Hence, the proposed project tries to tackle this issue through awareness campaigns, in support of the inclusivity initiative to promote safe internet for diverse user backgrounds.
Problem Statement 2: Internet users are not aware of specific patterns or terms that can point directly to a cyberbullying attempt. Hence, this project has worked an optimized deep learning-based detection tool to recognize any cyberbullying attempt at the early stage, which can also serve as an educational support tool to raise awareness of cyberbullying to promote safe internet space.
Project Implementation Narrative
This project was undertaken to provide awareness on cyberbullying issues, which is a critical factor in ensuring internet inclusivity. There are two main concerns, which are a lack of awareness of cyberbullying and also lack of tools to help identify the cyberbullying incidents. The first objective is to deliver awareness campaigns to various stakeholders.
We have managed to achieve this goal by exceeding the early plans of providing a total of 6 early sessions, whereby we have delivered 17 sessions to university, secondary, and primary school students (please refer to 17 Series of Awareness Campaigns on Cyberbullying.pdf attachment). We have put more emphasis on primary school students since they are the most vulnerable internet users. It is observable that many of them do not realize the categories of cyberbullying, whereas the majority of them believe cyberbullying is only if you post negative comments on the internet.
They do not realize that isolation attempts by exclusion from the online group as well as online impersonation are also some examples of cyberbullying. Moreover, most of them do not know how to react if they are facing cyberbullying attempts. Hence, we have targeted the younger generation and we must modify our presentation approach to suit the targeted audience, which is considerably different when giving an awareness campaign to adults. We have also written to a newspaper to highlight this issue and it was published on 15 February 2023 (please refer to Newspaper Article.pdf).
Apart from data, we have also worked on the early design of deep learning artificial intelligence classifiers to identify cyberbullying incidents based on major factors such as age, gender, ethnicity, and religion. Furthermore, we have also provided training initiatives to the team members by providing workshops, which were held in Malaysia and Indonesia, mostly focusing on deep learning technology. We have also developed various materials to support the cyberbullying awareness campaign in three languages, 1) Bahasa Malaysia (please refer to Awareness Campaign Presentation Slides in Bahasa Malaysia.pdf attachment), 2) English (please refer to Awareness Campaign Presentation Slides in English.pdf attachment), and 3) 1) Bahasa Indonesia (please refer to Pamphlet Cyberbullying Awareness 1 and 2 in Bahasa Indonesia attachment). As for the continual improvement exercise, we interviewed the audience through stratified sampling to get feedback on our awareness campaigns.
At the moment, we have optimized the deep learning classifier and we have sent the manuscript to Heliyon for the review process (please refer to Cyberbullying Detection using Attention-based Deep Neural Networks.pdf). The algorithm uses text-based input that focuses on five classes of major cyberbullying incidents. The optimal base architecture is first determined by analyzing a set of combined compact layers for a classification network. It is found that one-layer CNN is the optimal model to extract features from a limited 280 characters of tweet compared to multiple convolutional layers. As the number of convolutional layers are added, the accuracy of the model decreases and the loss of the model increases. The decrease in accuracy when more convolutional layers are added may be attributed to the increasing complexity of the model and causing it to overfit, given the limited number of text characters in a tweet. Having a model that is too complex given the size of input data may cause the model to start memorizing the training data as compared to learning the essence and generalizing the patterns from it. Then, an attention mechanism is embedded into the base network to put more emphasis on unique terms that represent various types of cyberbullying. Moreover, it allows the network to focus on key phrases and important features.
The first configuration of the attention mechanism that needs to be optimized is the number of encoder and decoder layers, which are varied from 2 to 4 layers. The numbers of encoder and decoder layers need to be finetuned according to the problem, which is in our case cyberbullying tweets. Increasing the number of encoder and decoder layers might improve the model’s accuracy but may also result in the model being more difficult to train optimally. Therefore, after running the simulations, the results show that 2 layers of encoder and decoder were deemed to be too simple, which resulted in the model being unable to capture good data representation. On the other hand, a set of 4 layers of encoder and decoder resulted in an increase in accuracy but it was too difficult to train optimally. The results show that the proposed attention mechanism-based network produced an accuracy of 0.8311 and a loss of 0.4880. In conclusion, the attention mechanism has managed to improve the performance of a text-based classifier in detecting cyberbullying incidences of various categories, which are due to religion, ethnicity, gender, and age.
We have also sent another manuscript to share our findings while delivering the awareness campaigns (please refer to Campaign on Awareness of the Negative Impacts of Cyberbullying.pdf). Finally, we have included technological tools in the latter awareness campaigns to help younger generations in identifying and overcoming cyberbullying incidents for primary and secondary school students. We have also openly shared our notes, activities, findings, and many more on a dedicated website. Any parties can access the materials and we do hope that it benefits users beyond our initial target groups. Finally, we have concluded our project through our closure workshop, which was held on 28 March 2024. In the closing workshop, several important points such as lessons learned and the impact of our activities are shared, so that we can identify any areas that need improvement as well as plan for our next steps.
Key Deliverables:
Beginning of Project
| Activity | Description | #Months |
|---|---|---|
| Dataset collection and cleaning. | To prepare a large dataset that contain cyberbullying incidences that will be used to construct the detection tool. | 4 |
Middle of Project
| Activity | Description | #Months |
|---|---|---|
| Awareness campaign on cyberbullying. | To provide seminars and awareness campaigns to stakeholders on the cyberbullying issue. | 8 |
End of Project
| Activity | Description | #Months |
|---|---|---|
| Stakeholder feedbacks and finetuning of the campaign and proposed system. | To collect user feedbacks on the developed cyberbullying system, so that the optimized version can be rolled out. | 3 |
| Promotion of the developed cyberbullying tool to the stakeholders. | To promote the usage of the developed cyberbullying detector as a safeguard tool to younger generation as well as parents. | 3 |
| Reporting and academic writing. | To prepare a manuscript and final report that discusses this project findings. | 1 |
Throughout the Project
| Activity | Description | #Months |
|---|---|---|
| Development of the deep learning-based cyberbullying system. | To develop an optimized deep learning network that is based on optimized convolutional neural network and recurrent neural network to recognize the cyberbullying incidences. | 12 |
Project Indicators
| Indicator | Status |
|---|---|
| Organize 1 (ONE) kickoff meeting/workshop and Hire 1 (ONE) Researcher by February 2023 (10%). | Completed |
| Organize 3 (THREE) activities concerning cyberbullying awareness by April 2023 (30%). | Completed |
| Develop 1 (ONE) deep learning-based cyberbullying classifier system by August 2023 (30%). | Completed |
| Organize 3 (THREE) cyberbullying awareness activities with the support of an AI tool by December 2023 (20%). | Completed |
| Organize 1 (ONE) closure meeting/workshop and publish 1 (ONE) academic writing by December 2023 (10%). | Completed |
Project Indicator Details
| Indicator: Organize 1 (ONE) kickoff meeting/workshop and Hire 1 (ONE) Researcher by February 2023 (10%). Status: Completed Start and End Dates: January 1, 2023 to February 28, 2023 Description: We will organize one kickoff meeting to facilitate and streamline the proposed activities, whereby one postdoctoral researcher will be hired to help with the project execution by February 2023. Baseline:This project is the starting point of collaboration between the involved researchers and no post-doctoral assistance has been hired. Activities: We have delivered a kickoff workshop and hired a post-doctoral assistance for the project. Outcomes: The kickoff meeting/workshop has been delivered successfully on 16 January 2023 and we have hired a post-doctoral since 1 February 2023. |
| Indicator: Organize 3 (THREE) activities concerning cyberbullying awareness by April 2023 (30%). Status: Completed Start and End Dates: March 1, 2024 to August 31, 2024 Description: We will develop the materials that will be used for cyberbullying awareness campaign and we will deliver at least three promotional activities to various stakeholders by April 2023. Baseline:No awareness campaigns have been deliverred. Activities: A series of 12 early awareness campaigns have been delivered (12 out of 17 awareness campaign, whereby the remainder campaigns are delivered after August 2023) Outcomes: Successfully delivered awareness campaigns, whereby the feedbacks are used to continually update the awareness campiagn's module. |
| Indicator: Develop 1 (ONE) deep learning-based cyberbullying classifier system by August 2023 (30%). Status: Completed Start and End Dates: January 1, 2023 to August 31, 2023 Description: We will develop a deep learning-based classifier, which can automatically identify the likelihood of cyberbullying attempts. The algorithm will be based on the Python platform, whereby convolutional neural network and recurrent neural network approaches will be developed by August 2023. Baseline:No deep learning algorithm has been designed for the automated cyberbullying detection. Activities: Various deep learning configurations have been designed, analyzed and optimized to produce the best detector based on English language. Outcomes: An optimized deep learning cyberbullying detector has been developed with an accuracy of ~82%. |
| Indicator: Organize 3 (THREE) cyberbullying awareness activities with the support of an AI tool by December 2023 (20%). Status: Completed Start and End Dates: September 1, 2023 to December 31, 2023 Description: We will organize at least three more cyberbullying awareness activities with the help of an AI support tool that was developed in the previous grant activity. A promotional website/online page will be designed by December 2023 to promote the tools and awareness activities. Baseline:The sharing sessions on cyberbullying at schools were executed without much help from technological tools. Activities: We have executed five more sharing sessions (to make a total of 17 sessions: 12+5) on cyberbullying awareness that include team activities among the students to help them clear their understanding of cyberbullying with the help of technological tools. Outcomes: Delivered active sharing- sessions on cyberbullying that included the usage of artificial intelligence tools to overcome cyberbullying. |
| Indicator: Organize 1 (ONE) closure meeting/workshop and publish 1 (ONE) academic writing by December 2023 (10%). Status: Completed Start and End Dates: February 28, 2024 to February 28, 2024 Description: We will organize one closure meeting or workshop to close the project, whereby we will publish all the findings in one academic writing by December 2023. Baseline:The team has not met for the final conclusion of the project. Activities: A closure workshop to conclude the project has been organized on 27 untul 29 February 2024 Outcomes: The closure workshop has allowed us to share our final findings that include lessons learnt, publications, reporting, and our next step. |
Project Review and Assessment
This project involves three objectives, which are 1) to conduct awareness campaigns on the cyberbullying phenomenon to diverse target groups, 2) to develop an intelligent cyberbullying detection tool based on an optimized deep learning network and 3) to integrate the cyberbullying detection tool in the promotional campaign. We have exceeded the first objective’s goal, in which we have delivered 17 sessions of awareness campaigns that cover various stakeholders in three countries; Malaysia, Indonesia, and Jordan.
We have also conducted two workshops to provide training to the project’s members; one workshop was held in Malaysia and one workshop was held in Indonesia. For the second objective, we have designed and tested the optimal cyberbullying classifier by using advanced deep learning methodologies, in which we detect cyberbullying incidences based on age, religion, ethnicity, and gender. For the third objective, we have executed four extensive workshops/sharing sessions with primary and secondary schools, whereby we have stressed the usage of artificial intelligence tools to identify cyberbullying incidents. We have also built a dedicated website at https://infocyberbully.wixsite.com/cyberbullying, so that our materials can be accessed by the public and the materials were prepared in English, Bahasa Malaysia and Bahasa Indonesia. To further solidify our findings we have also submitted two academic papers for review, whereby one focusing on the technical aspect (please refer to the Cyberbullying Detection using Attention-based Deep Neural Networks.pdf attachment), while one focusing on the community aspect (please refer to the Campaign on Awareness of the Negative Impacts of Cyberbullying.pdf attachment).
We believe that through our project, many students have benefited from our awareness campaign on the cyberbullying issue and we have promoted a safer way to use the internet, which directly increases the inclusivity of internet usage. At the end of some sessions, we conducted short interviews to get some feedback from the stakeholders, whereby we received good feedback in increasing the awareness of cyberbullying among students coming from various ages, races, ethnicities, and religions.
From the awareness sessions, many students are not aware of the danger of cyberbullying. Furthermore, most of them did not know what to do if they were facing cyberbullying attempts. The students also did not realize that there is a law in which bullies can be charged when they attempt to cyberbully others. Therefore, we believe that it is important to increase the promotional aspects of cyberbullying awareness, not just to the school students, but also to the university students as well as the public. We have uploaded all our materials to the online platform either existing media social or a dedicated website, as we believe that our materials can benefit many stakeholders.
Furthermore, we have received many good feedback from the teachers, hoping that the awareness campaign or any related cybersecurity issues can be expanded, which are crucial issues in the era of IR 4.0. The younger generations need to be equipped with the knowledge and tools to prevent them from being the victim of cyberbullying or any other cybersecurity threats. As for enhancing the technical capacity, the students and postdoctoral researchers who are directly involved with the project implementation have shown considerable growth in understanding deep learning technology. We have prepared the TWO journal manuscripts detailing 1) our findings on how deep learning artificial intelligence techniques can help identify cyberbullying incidences and 2) our findings while delivering the awareness sessions.
Apart from that, we have also learned an important lesson through our project engagement on how to deliver awareness campaigns to audiences of different ages. Generally, we are more familiar with giving talks to adults such as university students and hence have little experience in dealing with school students. We have modified our approach through the various sessions, by learning from our early session experiences and also from the teachers themselves. It is a unique experience when dealing with a large crowd of young age.
This project has benefited our project team members a lot, in which we have executed two workshops one in Malaysia and one in Indonesia to provide technical knowledge to the team members. Besides, by joining the APNIC mailing group, I have received a lot of information on various academic opportunities. Indeed, I have joined the Internet Governance Forum and am part of the team that has drafted the policy networks on artificial intelligence for the 2023 report and currently, I am involved in drafting the 2024 report.
Diversity and Inclusion
The main goal of the proposed project is to create awareness of cyberbullying and develop an intelligent tool to detect it. We strongly believe cyberbullying is one of the blocking factors for better internet inclusivity among diverse users, as such by reducing the cyberbullying incidents, a more inclusive internet space can be achieved. Hence, a more inclusive environment promotes diversity among internet users who come from various backgrounds, whereby they can feel safe to utilize the internet. The detection tool aims to identify cyberbullying incidents based on four major factors, which are age, gender, ethnicity, and religion, and hence, the project indirectly promotes better inclusion and gender equality among internet users. With regards to the team member aspect, the team is well balanced in terms of gender with two male and two female researchers. Statistically, the team consists of a balanced participation in terms of gender, ethnicity, and religion.
- Country: Malaysia (Dr. Zulkifley, Dr. Tan, Dr. Othman) and Indonesia (Dr. Muharar)
- Organization: Universiti Kebangsaan Malaysia (Dr. Zulkifley, Dr. Tan), Universiti Teknikal Malaysia Melaka (Dr. Othman) and Unviersitas Syiah Kuala (Dr. Muharar)
- Gender: Male (Dr. Zulkifley, Dr. Muharar), and female (Dr. Tan, Dr. Othman)
- Race: Majority in Malaysia (Dr. Zulkifley, Dr. Othman), Minority in Malaysia (Dr. Tan), Minority in Indonesia (Dr. Muharar)
- Religion: Islam (Dr. Zulkifley, Dr. Othman, Dr. Muharar) and Buddhism (Dr. Tan)
Furthermore, our target audience for all 17 sharing sessions that were held at several primary schools, secondary schools and university students come from various backgrounds of races, genders, and religions. We have also taken feedback videos on the awareness activities, whereby the stakeholders come from diverse backgrounds to ensure the inclusivity of the projects. Furthermore, there are three main languages involved in delivering the projects, which are English, Bahasa Malaysia, and Bahasa Indonesia. For the online tool, we have focused only on English-based detectors due to the limitation of project duration (one year project). We have shared our project activities with the university, and it produces a spilled-over effect in reducing cyberbullying, in order to increase inclusivity of internet usage. The findings as well as all developed awareness materials can be found at https://infocyberbully.wixsite.com/cyberbullying and we hope our findings can help a larger community, not just the targeted direct beneficiaries.
Project Communication
This project contains a number of activities that are related to promotion, whereby the developed cyberbullying tool helps in raising the awareness level of this issue among various target groups. Hence, several promotion strategies have been executed to ensure a wider set of users can be reached. We have published an article on cyberbullying awareness in the local newspaper as part of the project communication strategy. The promotion campaigns have been delivered to various schools and universities through face-to-face and online engagement. Furthermore, in our pamphlet and slides, we have acknowledged the ISIF Asia grant and also our respective universities as part of the promotional and communication strategies. Due to the various promotional strategies, this project has provided positive impacts on diverse groups of users. Besides, each of the team members has participated in the promotional campaign, especially to their own university students and surrounding communities, whereby our team is comprised of three different organizations. The first and third objectives require assessment feedback from the participants in order to measure the effectiveness of the suggested activities. We have recorded feedback videos through randomized sampling on the audiences and ensured the stakeholders cover diverse races, religions, and ethnicities. The feedback has been used to improve the effectiveness of the promotional campaign as well as the features of the developed cyberbullying detection tool. Since we are proposing to detect cyberbullying incidents because of several factors including age, gender, religion, and ethnicity, the promotional campaigns have tried to reach wide targeted users so that internet inclusivity among various users can be achieved by reducing cyberbullying cases. Besides that, we would like to highlight also that face-to-face sessions are more effective in dealing with school students, because they allow us to control the crowd better and enable us to convey the message effectively.
Impact Story
Throughout the 17 series of sharing sessions and workshops with the primary, secondary, and university students, we have realized that Malaysians and Indonesians need a more comprehensive awareness campaign on cyberbullying. In one of the shared stories by one of the participants, whereby she explained her side of the story and we found out that she ticked all the general symptoms of a cyberbullying victim, but because of lack of support, she still denied the incidents and treat the bully incident as just playing jokes among friends. In another story, a school counselor has specifically highlighted the consequence of cyberbullying to the participants, because she said that one of the students is actually a cyberbullying victim, and she wants us to highlight the negative consequence of cyberbullying to the bullies in general. We have also conducted video-recorded feedback from various school students and many of them agreed that they had low awareness of cyberbullying before the workshops and they have benefited a lot from our sharing session. Please refer to the Transcribed Feedbacks from Participants.pdf attachment for full survey feedback from the students.
We strongly believe that our sharing sessions have had a positive impact on the participants and a few schools have shared their feedback through appreciation letters, in which the teachers have shared some of our materials with other students who have not attended our sessions. One of the lessons learned that we would like to highlight is when we delivered the session to Sekolah Menengah Kebangsaan Bandar Baru Bangi and we just happened to receive participants that are part of Peer Mentoring Prefects. This set of prefects is tasked to help mentor their friends, who have issues with their studies and they do highlighted that bullying either through physical or online, is one of the major issues that they are facing. We received feedback from the teachers that the prefects have benefited from the workshops and they have been tasked to create short notes to help their friends, which we believe should be beneficial because of the advice from peers.
In terms of university impact, we have sent this project for rating approval. In Malaysia, the universities are rated every year, whereby one of the criteria is community activities. Hence, we have sent our application with regard to this project for a five-star rating, which is the highest rating possible.
In terms of technological advancement, we have submitted a journal manuscript detailing our design of the advanced deep artificial intelligence cyberbullying detector to a high-impact publication (Heliyon), and also another journal manuscript detailing our findings while delivering the awareness campaigns. The developed system is based on text input that concerns five major classes of cyberbullying. The optimal base architecture consists of one layer of convolutional neural networks with two layers of connected components, whereby the hyperparameters have been optimized and compared with the other state-of-the-art models. An attention mechanism has been embedded into the base network to put more emphasis on unique terms that represent various types of cyberbullying. The results show that the proposed attention mechanism-based network produced an accuracy of 0.8311 and a loss of 0.4880. In conclusion, the attention mechanism has managed to improve the performance of a text-based classifier in detecting cyberbullying incidences of various categories, which are due to religion, ethnicity, gender, and age.
Project Sustainability
With regards to our awareness campaign materials, we have provided them for free through a dedicated website, which can be accessed at https://infocyberbully.wixsite.com/cyberbullying. We have also interviewed the stakeholders and we received good feedback on how to identify and help the friend who might be facing the cyberbullying threat. It is important to teach the students at a young age so that they can share the information with their friends, thus reducing cyberbullying incidences. We also realized that cyberbullying is closely related to cybersecurity, whereby some of the bullying incidents can be prevented if internet users know how to browse the internet safely. Continuing from this concern, we are aiming to scale up this project to also include cybersecurity topics that aims to target Peer Mentoring Prefects for a bigger impact. For this project also, we have received a small amount of top-up grant of RM 13,194 (~ USD 3,000) to ensure the project can be delivered smoothly starting 15 May 2023.
Project Management
In this project, Dr. Zulkifley has served as the artificial intelligence expert, while Dr. Othman has provided expertise on cyberbullying. While Dr. Muharar has provided expertise on networking components and Dr. Tan helps in campaign visualization. Furthermore, this project has hired one postdoctoral researcher to help with the overall project implementation. Our project team has delivered various activities by using the shared materials that we have gathered during the early stage of the project. We have learned from each other and any lessons learned are used to continually improve the project implementation. Hence, our modules are polished from time to time while delivering all 17 sharing sessions. To further polish our team member’s skills, we have also organized two capacity-building workshops, one in Malaysia and one in Indonesia to ensure the benefit to all team members.
This project was closely monitored by the UKM’s Research Management Center which requires us to provide periodical updates of the project's progress. Therefore, any issue that may hinder the project's progress can be tracked effectively and reduce the risk from various factors. It is also worth to note that the project’s progress and budget were tracked through an effective online system (https://smp.ukm.my/). Besides that, the risk of technical, financial, and timing aspects of the project is considered as low as detailed out below:
- Technical: The lead researcher has in-depth expertise in deep learning technology, which he has successfully applied to various research problems that include biomedical diagnosis, finance-related prediction, remote sensing, and computer vision applications.
- Financial: The risk in terms of monetary is low as all the requested items can be sourced from local companies and hence, limit the problem of import tax issues.
- Timing: The team has agreed to use the online platform as the primary communication channel, which limits the delay caused by a face-to-face meeting. Besides, we have prior knowledge from executing the small grant effort, and thus, the timing delay is expected to be minimal.
We have submitted this project the for Malaysian Research Assessment (MyRA) star rating, in which we have applied for the full five-star acknowledgment. The assessment is still being processed and hopefully, we can get the five-star rating. Generally, this project has raised awareness on the issue of cyberbullying not just among our own respective organizations but also to many stakeholders, especially when we made all our materials freely accessible online at https://infocyberbully.wixsite.com/cyberbullying. We also believe that Universiti Kebangsaan Malaysia has consolidated our strength in this issue and we would like to take it a notch further through our newly proposed scale-up project.
Project Recommendations and Use of Findings
When dealing with the school students, active participation is crucial in order to control the crowd, so that our message can be conveyed effectively. Alternatively, it is easier to give a talk to university students as they are more disciplined and can give better attention during the talk. It is also surprising that many students are not aware of some cases of cyberbullying such as exclusion incidents from a certain online platform or posting shameful things about the victim. Some of the incidents are treated as teasing among the friends, and they do not realize how it can affect the victim. Therefore, it is important to provide awareness to the younger generations so that they can avoid from being a victim of cyberbullying. In fact, they are also taught some cybersecurity knowledge such as having a safe and secure password so that the bully can not access the victim's online account and avoiding the online impersonation threat.
Bibliography
1. Syamsul Azlan Saleh & Mohd Asyraf Zulkifley, “Segera Bentang Undang-Undang Khusus Tangani Kes Buli Siber”, Berita Harian, 15 February 2023.
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License 
