Multiplicative Vector Fusion Model for Detecting Deepfake News in Social Media

Round 1

Reviewer 1 Report

Review of the Manuscript applsci – 2238902-  „Multiplicative Vector Fusion Model for Detecting Deepfake News in social media” for the Applied Sciences.

General Comments

            From my point of view, it is a very interesting topic and simultaneously it seems that to the best of my knowledge is the first empirical research which study the role of the Multiplicative Vector Fusion Model for Detecting Deepfake News in social media.

The paper consists of the following sections: Introduction, Background and Literature Survey, Proposed Methodology, Results and Analysis and Conclusion and Future Work.

However, I find some recommendations:

1.       It would be very useful to add in the "Introduction" section the purpose, objectives and hypothesis of the research.

2.       The conclusions section should be developed more.

3.       I recommend the authors to make a complete descriptive analysis and to include a series of indicators and tests such as standard deviation, Jarque-Berra, Kurtosis, probabilities, etc., and the number of observations taken in the sample.

4.       It is very important that the authors present the correlation matrix and the ovariance matrix and explain the results obtained.

5.       I think that the authors should present the VIF test to verify heteroskedasticity and also to verify endogeneity.

6.       At the same time, the authors must explain why they chose cross-section with fixed effects and not with random effects. That is why the authors must do the Hausman test.

7.       Also,  we consider the literature is not enough and that is why, we recommend the authors to refer to other recent works indexed in Web of Science, Scopus, Emerald, Cambrige, and of course MDPI Journals. We suggest that the authors cite papers published in MDPI journals and Web of Science Journals, such as:

      Batrancea, L., Rathnaswamy, M.M., MI Rus, Tulai H., (2022), Determinants of Economic Growth for the Last Half Century: A Panel Data Analysis on 50 Countries, Journal of the Knowledge Economy. https://doi.org/10.1007/s13132-022-009944-9.

The conclusions must be expanded with possible economic policy implications of the research undertaken.

All in all, I consider that the paper must be improved. As a result, the article can be published in the prestigious Applied Sciences journal after minor revisions.

Author Response

Response to Reviewer 1 Comments

Firstly, I sincerely thank the reviewer for the thoughtful comments and constructive suggestions on the current study that will improve the manuscript’s comprehensibility. For quick reference, the portions revised in the manuscript are highlighted for your kind reference and consideration. I thank the reviewer for their valuable time in reviewing and recommending the corrections in the manuscript.

Our study has included the usage of the Grammarly app to enhance the quality of the English used.

Reviewer Comment 1:   It would be very useful to add in the "Introduction" section the purpose, objectives and hypothesis of the research.

Response 1: As per the reviewer's suggestions we have modified the introduction section including purpose and objectives.(Refer to pages 1,2)

There has been a lot of discussion about fake news during the 2016 US election, but the topic is not new. Media outlets, journalists, and editors usually follow a strict code of behaviour when giving out the news. The internet introduced a new way of consuming, publishing, and sharing information in the late twentieth century. Many people turn to social media for news these days. Around half of the world's population uses social media. In this way, fake news can sometimes be difficult to discern. Regarding news dissemination, social media sites and networks offer many advantages, including instant access to information, free distribution, and no time constraints. There needs to be more regulation for these platforms.

Using social media to disseminate false information undermines trust in the news ecosystem, damages individual and organizational reputations, and causes fear among the general public, all of which threaten society's stability. Since fake news uses terminology similar to real news, it is very difficult to distinguish fake news from real news. Fake news exists to instil confidence in the public. In order to prevent rumours, identity theft, a lack of authenticity and confidentiality, and fake profiles across online platforms, there is a heightened need to deal with the spread of false data across online platforms we used the RoBERTa base model as the foundation for our fake news detector model.

The following list of contributions is made to this article.

• Initially, we performed data-pre-processing and sentence generation from deep fake news datasets with the T5 model.
• We designed a fine-tuned Roberta model to detect deep fake and real news effectively with optimal parameters.
• To classify deep fake news from real news on social media datasets, we proposed a transfer learning-based multiplicative vector fusion (TL-MVF) model.
• The proposed TL-MVF model has been tested and validated on real-time and benchmarked datasets.
• We evaluated the TL-MVF model by taking into consideration Accuracy, Precision, Recall, AUC, and F-score.
• Finally, the proposed TL-MVF model outperformed the existing baseline framework.

Reviewer Comment 2: The conclusions section should be developed more.

Response 2: As per the reviewer's suggestions we have modified the conclusion section of the paper is available on page 21,22.

During the past decade, transformers' effectiveness has grown dramatically in NLP applications. An extensive range of social media applications is used to validate current deep-learning models for detecting fake news. However, models couldn’t perform much in real-world applications if they are not generalizable. Improving performance in clinical contexts requires training in a universal rather than a specialized model. Fake news classification and detection is a vital area in the current social media world, and the paper discusses the work that has been done in this regard. TL-MVF is a novel model proposed in this paper for detecting and categorizing fake news in social media. The TL-MVF model involves a series of sub-processes namely data precession, the T5 model for text-text sentence generation, the RELU activation function for optimizing the results, for fake news detection we fine-tuned the RoBERTa model, and the Adam-based hyperparameter tuning. Based on the titles of the long texts, a multiplicative fusion-based classification technique has been proposed for classifying news as fake or real. The experimental results of the TL-MVF model are performed utilizing one benchmark dataset and two real-time datasets, and the outcomes are investigated under distinct measures. The comparison study highlighted the enhanced performance of the TL-MVF model on existing approaches. TL-MVF is therefore an effective model for detecting and categorizing fake news on social media. For the purpose of this study, we simply considered text to identify bogus news on social media. As an added benefit of this research, the cyber cell of the police department can use it to adopt appropriate measures and methods for dealing with fake data, which will lower levels of crime and improve the quality of life for everyone. As the system was designed to analyze text data, its only limitation is that it only produces results for text data; however, in the future, it can be extended to include images alongside text to create broad and heterogeneous analysis results. Going forward, it is possible to study and test the given model in a variety of transformer models to improve the overall classification performance on fake images and video data. In addition, we plan to fine-tune the RoBERTa and subsequent layer hyperparameters and analyze their layered process in depth.

Reviewer Comment 3: I recommend the authors make a complete descriptive analysis and include a series of indicators and tests such as standard deviation, Jarque-Berra, Kurtosis, probabilities, etc., and the number of observations taken in the sample.

Response 3: As per the reviewer's suggestions we evaluated our approach by contrasting it to the base-line other research models such as 3HAN, HAN, CNN-RNN, CNN-LSTM, BERT-NLI, Fake BERT models which are seen in Section 4.3 and in our future work we will implement the tests you mentioned.

Reviewer Comment 4: It is very important that the authors present the correlation matrix and the covariance matrix and explain the results obtained.

Response 4: As per the reviewer's suggestions we provided a confusion matrix and loss function representations in Section 4.3 for explaining the results seen on (Page16)

Reviewer Comment 5: I think that the authors should present the VIF test to verify heteroskedasticity and also to verify endogeneity.

Response 5: As per the reviewer's suggestions what we define as VIF measures the strength of the correlation between the independent variables in regression analysis which can cause problems for regression models that’s why letting us not choosing this test but in future works, we will try to implement this.

Reviewer Comment 6: At the same time, the authors must explain why they chose a cross-section with fixed effects and not with random effects. That is why the authors must do the Hausman test.

Response 6: As per the reviewer's suggestions we mentioned why we are with a cross-section with fixed effects and not with random effects in Section 4.5 which is observed on (Page 20)

Overfitting is a problem that must be addressed with special care when it comes to transferring learning networks that have been accomplished by the proposed model. To discourse the overfitting problem, the model training along with validation performance is measured. The results of this monitoring are collectively presented across various optimizers in Fig. 11. When comparing the various optimizers, it is important to keep in mind that the value range on the y-axis varies from one to the next. This allows for a more accurate depiction of the variations. In addition to that, the findings that were achieved from the experiments with various optimizers and learning settings are described in detail.

Reviewer Comment 7: Also, we consider the literature is not enough and that is why, we recommend the authors refer to other recent works indexed in Web of Science, Scopus, Emerald, Cambridge, and of course MDPI Journals. We suggest that the authors cite papers published in MDPI journals and Web of Science Journals, such as:

Response 7: As per the reviewer's suggestions we added recent references which are suitable for the paper and are seen on pages 22,23.

Reviewer 2 Report

This paper proposed a multiplicative vector fusion model for classifying fake news from real news efficiently. It is interesting and novelty. 

If the following points could be improved, would be better:

(1)         Line 18, as T5, please give abbreviation full words at the beginning.

(2)        As Figure2, it is better to explain the details of low-level lexical features, syntactic features 

Author Response

Response to Reviewer 2 Comments

Firstly, I sincerely thank the reviewer for the thoughtful comments and constructive suggestions on the current study that will improve the manuscript’s comprehensibility. For quick reference, the portions revised in the manuscript are highlighted for your kind reference and consideration.I thank the reviewer for their valuable time in reviewing and recommending the corrections in the manuscript.

Our study has included the usage of the Grammarly app to enhance the quality of the English used.

Reviewer Comment 1: Line 18, as T5, please give the abbreviation full words at the beginning.

Response 1: As per the reviewer's suggestions we have modified and added full abbreviations at the beginning. (Refer to page 1)

To generate the sentences, the T5, or Text-to-Text Transfer Transformer model was employed for data cleaning and feature extraction

Reviewer Comment 2: As in Figure 2, it is better to explain the details of low-level lexical features and syntactic features.

Response 2: As per the reviewer's suggestions we have modified and explained the low-level lexical features, and syntactic feature's importance which are available on page 8,9.

Fake news has been identified by many studies using feature-based classification. Using textual characteristics, it is easy to detect false information. A few of the features are discussed below:

• An essential aspect of a text's semantics is its meaning (semantics). In this way, the data is transformed into meaningful patterns.
• Word frequency and uniqueness are calculated using lexical features in the TF-IDF vectorization. Hashtags, pronouns, and punctuation are some of the lexical features.
• Syntactic features are generated by speech tags and various components from a parse tree, whereas lexical features are the target words with unigrams, bigrams, and surface forms.

Reviewer 3 Report

Authors must justify novelty of the proposed work. Novelty is missing.
Authors must add more recent and relevant references
authors must add a table of comparison to showcase how/where their proposed approach is better.
Authors must discuss limitations of the proposed approach.
Authors must discuss possible future development based on these limitations

Author Response

Response to Reviewer 3 Comments

Firstly, I sincerely thank the reviewer for the thoughtful comments and constructive suggestions on the current study that will improve the manuscript’s comprehensibility. For quick reference, the portions revised in the manuscript are highlighted for your kind reference and consideration. I thank the reviewer for their valuable time in reviewing and recommending the corrections in the manuscript.

Our study has included the usage of the Grammarly app to enhance the quality of the English used.

Reviewer Comment 1: Authors must justify the novelty of the proposed work. Novelty is missing.

Response 1: As per the reviewer's suggestions we justified the novelty of the proposed work. (Refer to page 2)

Using social media to disseminate false information undermines trust in the news ecosystem, damages individual and organizational reputations, and causes fear among the general public, all of which threaten society's stability. Since fake news uses terminology similar to real news, it is very difficult to distinguish fake news from real news. Fake news exists to instil confidence in the public. In order to prevent rumours, identity theft, a lack of authenticity and confidentiality, and fake profiles across online platforms, there is a heightened need to deal with the spread of false data across online platforms we used the RoBERTa base model as the foundation for our fake news detector model.

The following list of contributions is made to this article.

• Initially, we performed data-pre-processing and sentence generation from deep fake news datasets with the T5 model.
• We designed a fine-tuned Roberta model to detect deep fake and real news effectively with optimal parameters.
• To classify deep fake news from real news on social media datasets, we proposed a transfer learning-based multiplicative vector fusion (TL-MVF) model.
• The proposed TL-MVF model has been tested and validated on real-time and benchmarked datasets.
• We evaluated the TL-MVF model by taking into consideration Accuracy, Precision, Recall, AUC, and F-score.
• Finally, the proposed TL-MVF model outperformed the existing baseline framework.

Reviewer Comment 2: Authors must add more recent and relevant references

Response 2: As per the reviewer's suggestions we added recent references which are suitable for the paper and are seen on pages 22,23.

Reviewer Comment 3: Authors must add a table of comparison to showcase how/where their proposed approach is better.

Response 3: As per the reviewer's suggestions we have compared and highlighted the comparative study of the TL-MVF model with other existing mechanisms' overall evaluation measures on the Fake & Real news dataset presented in Tables 6,7, and 8 on Pages 17,18 and 19.

Reviewer Comment 4: Authors must discuss the limitations of the proposed approach

Response 4: As per the reviewer's suggestions we have discussed the limitations of the proposed approach (Page 5)

The following list of limitations is observed over the literature survey.

• Detecting deep fake news is a challenge because of the lack of a benchmarked labelled dataset with actual truth labels and complete information space.
• False news has become increasingly widespread and difficult to detect in today's environment.
• The most challenging aspect of spotting fake news is doing it early on. The lack of data to train detection models is another issue with fake news detection.
• To identify false news, it is necessary to have a solid awareness of specific authors, entities, and the relationship that exists between each word in a lengthy text.
• To overcome the above-aforementioned issues, we proposed a transfer learning-based multiplicative vector fusion (TL-MVF) model has been proposed and implemented.

Reviewer Comment 5: Authors must discuss possible future development based on these limitations

Response 5: As per the reviewer's suggestions we have added future development of the proposed approach (Pages 21,22)

Conclusion and Future Work

During the past decade, transformers' effectiveness has grown dramatically in NLP applications. An extensive range of social media applications is used to validate current deep-learning models for detecting fake news. However, models couldn’t perform much in real-world applications if they are not generalizable. Improving performance in clinical contexts requires training in a universal rather than a specialized model. Fake news classification and detection is a vital area in the current social media world, and the paper discusses the work that has been done in this regard. TL-MVF is a novel model proposed in this paper for detecting and categorizing fake news in social media. The TL-MVF model involves a series of sub-processes namely data precession, the T5 model for text-text sentence generation, the RELU activation function for optimizing the results, for fake news detection we fine-tuned the RoBERTa model, and Adam-based hyperparameter tuning. Based on the titles of the long texts, a multiplicative fusion-based classification technique has been proposed for classifying news as fake or real. The experimental results of the TL-MVF model are performed utilizing one benchmark dataset and two real-time datasets, and the outcomes are investigated under distinct measures. The comparison study highlighted the enhanced performance of the TL-MVF model on existing approaches. TL-MVF is therefore an effective model for detecting and categorizing fake news on social media. For the purpose of this study, we simply considered text to identify bogus news on social media. As an added benefit of this research, the cyber cell of the police department can use it to adopt appropriate measures and methods for dealing with fake data, which will lower levels of crime and improve the quality of life for everyone. As the system was designed to analyze text data, its only limitation is that it only produces results for text data; however, in the future, it can be extended to include images alongside text to create broad and heterogeneous analysis results. Going forward, it is possible to study and test the given model in a variety of transformer models to improve the overall classification performance on fake images and video data. In addition, we plan to fine-tune the RoBERTa and subsequent layer hyperparameters and analyze their layered process in depth.

Reviewer 4 Report

Title: Multiplicative Vector Fusion Model for Detecting Deepfake 2 News in social media

The authors have developed a multiplicative vector fusion model for detecting Deepfake news in social media. In a way, the practical analysis used supports the presented framework due to my own observation, and the paper is also relevant to this conference. However, the author has to look into the following concerns:

1.      The motivation and contribution of this paper should be stated more clearly in the abstract to better understand from the beginning of the study. Authors are advised to be precise in the abstract, and structure your abstract as follows- 1) Background 2) Aim/Objective 3) Methodology 4) Results 5) Conclusion. Write 2-4 lines for each and merge everything in one paragraph (200-300 Words) without any subheading.

1.      The Introduction section should be improved by adding more recent works in this area and providing a more accurate and informative literature review with the pros and cons of the available approaches and how the proposed method is different comparatively.

2.      The related work section is very small, an updated and complete literature review should be conducted and should appear in section 2- Related Work. Some latest papers which studied similar effects problems can be discussed to help the readers. Avoid the explanation of common concepts like Transformer learning, Self-attention, Transformers etc. try to review and cite this related paper

a.      Awotunde, J. B., Jimoh, R. G., Imoize, A. L., Abdulrazaq, A. T., Li, C. T., & Lee, C. C. (2022). An Enhanced Deep Learning-Based DeepFake Video Detection and Classification System. Electronics, 12(1), 87.

3.      The figures presented are not too clear, authors should please work on their figures for better quality. For examples figure 1 and 2 is not readable.

4. Subsections 4.1 to 4.3 should be moved to the proposed methodology section. They are part of the methodology, not results and discussion. Section 4 should start from subsection 4.4 forward.

5.      There should be a discussion section to give a general discussion on the results gotten from the proposed model and explain in detail the comparative analysis of the model with existing methods.

6.      The authors need to further emphasize their contributions and relate with the results obtained as is why the adopted methods used are better than the others.

7.      Furthermore, the authors fail in explaining the details of their approach in a clear manner. The overall description and details regarding how the authors derive the result are not clear. The flowchart details are given without sufficient explanation and only a few variables appearing in them are addressed, leaving a large part of those obscure.

8.      The author seems to disregard or neglect some important findings in the results that have been achieved in the paper. So elaborate and explain the results in more detail.

9.      It would be interesting that the author explains the limitations of the present study to help other authors for future studies. Mention the future scope of your present works.

10. Although the English is generally quite good, there are quite a few minor grammatical errors, and a careful read-through is needed to eliminate these errors. The spelling mistake should be corrected by reading through the manuscript.

I appreciate the style of presentation of this paper, but the author needs to incorporate the above-mentioned points for a better and possible publication in the forthcoming conference. I, therefore, recommend a major revision.

Author Response

Response to Reviewer 4 Comments

Firstly, I sincerely thank the reviewer for the thoughtful comments and constructive suggestions on the current study that will improve the manuscript’s comprehensibility. For quick reference, the portions revised in the manuscript are highlighted for your kind reference and consideration. I thank the reviewer for their valuable time in reviewing and recommending the corrections in the manuscript.

Our study has included the usage of the Grammarly app to enhance the quality of the English used.

Reviewer Comment 1: The motivation and contribution of this paper should be stated more clearly in the abstract to better understand from the beginning of the study. Authors are advised to be precise in the abstract, and structure your abstract as follows- 1) Background 2) Aim/Objective 3) Methodology 4) Results 5) Conclusion. Write 2-4 lines for each and merge everything in one paragraph (200-300 Words) without any subheading.

Response 1: As per the reviewer's suggestions we structured the abstract by merging all the above-mentioned things. (Refer to page 1)

In the digital age, social media platforms are becoming vital tools for generating and detecting deep fake news due to the rapid dissemination of information. Unfortunately, today, fake news is being developed at an accelerating rate that can cause substantial problems, such as early detection of fake news, a lack of labelled data available for training, and identifying fake news instances that still need to be discovered. Identifying false news requires an in-depth understanding of authors, entities, and the connection between words in a long text. Unfortunately, many Deep Learning (DL) techniques have proven ineffective with lengthy texts to address these issues. This paper proposes a TL-MVF model based on transfer learning for detecting and generating deep fake news in social media. To generate the sentences, the T5, or Text-to-Text Transfer Transformer model was employed for data cleaning and feature extraction. In the next step, we designed an optimal hyperparameter Roberta model for effectively detecting fake and real news. Finally, we propose a multiplicative vector fusion model for classifying fake news from real news efficiently. A real-time and benchmarked dataset has been used to test and validate the proposed TL-MVF model. For the TL-MVF model, F-score, Accuracy, Precision, Recall, and AUC were performance evaluation measures. As a result, the proposed TL-MVF performed better than existing benchmarks.

Reviewer Comment 1: The Introduction section should be improved by adding more recent works in this area and providing a more accurate and informative literature review with the pros and cons of the available approaches and how the proposed method is different comparatively.

Response 1: As per the reviewer's suggestions we added recent works with pros and cons. (Refer to pages 2 and 3)

There has been a lot of discussion about fake news during the 2016 US election, but the topic is not new. Media outlets, journalists, and editors usually follow a strict code of behaviour when giving out the news. The internet introduced a new way of consuming, publishing, and sharing information in the late twentieth century. Many people turn to social media for news these days. Around half of the world's population uses social media. In this way, fake news can sometimes be difficult to discern. Regarding news dissemination, social media sites and networks offer many advantages, including instant access to information, free distribution, and no time constraints. There needs to be more regulation for these platforms.

Using social media to disseminate false information undermines trust in the news ecosystem, damages individual and organizational reputations, and causes fear among the general public, all of which threaten society's stability. Since fake news uses terminology similar to real news, it is very difficult to distinguish fake news from real news. Fake news exists to instil confidence in the public. In order to prevent rumours, identity theft, a lack of authenticity and confidentiality, and fake profiles across online platforms, there is a heightened need to deal with the spread of false data across online platforms we used the RoBERTa base model as the foundation for our fake news detector model.

The following list of contributions is made to this article.

• Initially, we performed data-pre-processing and sentence generation from deepfake news datasets with the T5 model.
• We designed a fine-tuned Roberta model to detect deep fake and real news effectively with optimal parameters.
• To classify deep fake news from real news on social media datasets, we proposed a transfer learning-based multiplicative vector fusion (TL-MVF) model.
• The proposed TL-MVF model has been tested and validated on real-time and benchmarked datasets.
• We evaluated the TL-MVF model by taking into consideration Accuracy, Precision, Recall, AUC, and F-score.
• Finally, the proposed TL-MVF model outperformed the existing baseline framework.

Reviewer Comment 2: The related work section is very small, an updated and complete literature review should be conducted and should appear in section 2- Related Work. Some latest papers which studied similar effects problems can be discussed to help the readers. Avoid the explanation of common concepts like Transformer learning, Self-attention, Transformers etc. try to review and cite this related paper

87. Awotunde, J. B., Jimoh, R. G., Imoize, A. L., Abdulrazaq, A. T., Li, C. T., & Lee, C. C. (2022). An Enhanced Deep Learning-Based DeepFake Video Detection and Classification System. Electronics, 12(1), 87.

Response 2: As per the reviewer's suggestions we have added data to Section 2 i.e; Background and Literature Survey by citing recent papers and we explained Transformer learning, Self-attention, Transformers etc to describe the techniques used in our work as on Page 5.

In conclusion, fake news can be identified from social media contexts, malicious user profiles, and user activities. However, despite their benefits, these approaches present several challenges as well. For instance, it is difficult to enumerate social contexts since that is a broad field. Furthermore, detection algorithms may need to be more effective due to a large amount of data and its incompleteness, noise, and unstructured nature.

Reviewer Comment 3: The figures presented are not too clear, authors should please work on their figures for better quality. For examples figure 1 and 2 is not readable.

Response 3: As per the reviewer's suggestions we have modified the figures in a readable way for everyone.

Reviewer Comment 4: Subsections 4.1 to 4.3 should be moved to the proposed methodology section. They are part of the methodology, not results and discussion. Section 4 should start from subsection 4.4 forward.

Response 4: As per the reviewer's suggestions we have moved the proposed methodology section 4.1 to 4.3 to Section 3 which is seen in (Pages 6,7 and 8)

Reviewer Comment 5: There should be a discussion section to give a general discussion on the results gotten from the proposed model and explain in detail the comparative analysis of the model with existing methods.

Response 5: As per the reviewer's suggestions we have discussed results obtained from the proposed model in Section 4.3 along with a comparison of different machine learning and deep learning techniques is done to assess the performance of fake news detection (Pages 15 to 21).

Reviewer Comment 6: The authors need to further emphasize their contributions and relate with the results obtained as is why the adopted methods used are better than the others.

Response 6: As per the reviewer's suggestions we have discussed our proposed method priority when compared with others which are seen in (Page 12)

To detect fake tweets on social media, we've implemented a Robustly optimized BERT technique (Roberta) that has been pre-trained. With our benchmarked data sets, we can fine-tune the RoBERTa model by replacing its last layer with a SoftMax layer. Better performance for the existing model is only attainable through careful hyperparameter fine-tuning. Hyperparameters in this model include the sequence length, linear layer choice, neuron count in each layer, optimizer, learning rate, minibatch size, and epochs. During training, the technique maintains the same weights as the pre-trained model. Aside from the variation in minibatch size, the final produced model with all hyperparameters is the same. This is because the two datasets are of different sizes. Besides capturing left-to-right text directions as well as right-to-left text directions, Roberta can also learn additional context information from a tweet. In comparison to earlier versions like BERT-base and BERT large, RoBERTa's improved performance can be attributed to the following changes:

• The RoBERTa model is pre-trained with 10 times more data and 8 times larger batch sizes.
• As opposed to character-level vocabulary techniques, the model used BPE (Byte-Pair-Encoding).
• NSP (Next Sentence Predicter) has been removed from the model.
• Changed the crucial parameters such as masking patterns applied dynamically, higher learning rates, etc.

Reviewer Comment 7: Furthermore, the authors fail in explaining the details of their approach in a clear manner. The overall description and details regarding how the authors derive the result are not clear. The flowchart details are given without sufficient explanation and only a few variables appearing in them are addressed, leaving a large part of those obscure.

Response 7: As per the reviewer's suggestions we have explained the Implementation details to derive the result in Section 4.2 on Page 15.

We used a high-end configuration system to implement the proposed model. All the implementations were done using the Linux I7 operating system, 16 GB DDR4 RAM, 1 TB SSD, and supports 32 GB GPU. The Anaconda navigator with Spyder environment and experimented on Kera’s. All the models and algorithms are programmed using Python language. The hyperparameters and their values utilized to build the model are shown in Table 5.

Reviewer Comment 8: The author seems to disregard or neglect some important findings in the results that have been achieved in the paper. So elaborate and explain the results in more detail.

Response 8: As per the reviewer's suggestions we have discussed all the obtained or achieved results with discussion in Section 4.3 on Pages (15 to 21).

Reviewer Comment 9: It would be interesting if the author explains the limitations of the present study to help other authors for future studies. Mention the future scope of your present works.

Response 9: As per the reviewer's suggestions we have discussed the limitations of the proposed approach (Page 5) and the future scope for this work is mentioned on Pages(21,22)

The following list of limitations is observed over the literature survey.

• Detecting deep fake news is a challenge because of the lack of a benchmarked labelled dataset with actual truth labels and complete information space.
• False news has become increasingly widespread and difficult to detect in today's environment.
• The most challenging aspect of spotting fake news is doing it early on. The lack of data to train detection models is another issue with fake news detection.
• To identify false news, it is necessary to have a solid awareness of specific authors, entities, and the relationship that exists between each word in a lengthy text.
• To overcome the above-aforementioned issues, we proposed a transfer learning-based multiplicative vector fusion (TL-MVF) model has been proposed and implemented.

Reviewer Comment 10:Although the English is generally quite good, there are quite a few minor grammatical errors, and a careful read-through is needed to eliminate these errors. The spelling mistake should be corrected by reading through the manuscript.

Response 10: As per the reviewer's suggestions we have read the entire manuscript and modified the grammatical errors.