Content uploaded by Anietie Ekong
Author content
All content in this area was uploaded by Anietie Ekong on Oct 17, 2024
Content may be subject to copyright.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
58 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
ABSTRACT: In the digital era, the proliferation of digital
content has intensified concerns over intellectual property rights
infringement, highlighting the need for robust copyright
protection solutions. This paper presents a software solution
designed to address these challenges by combining advanced
algorithms with intuitive user interfaces for effective copyright
enforcement. Central to the software’s functionality is the Most
Significant Bit (MSB) embedding technique, which allows users to
imperceptibly embed copyright or trademark information into
digital images. This method modifies the MSB of pixel values to
encode protection data while maintaining the visual integrity of
the images. In the detection phase, the software employs Deep
Convolutional Neural Networks (DCNN) to identify instances of
unauthorized use or copyright infringement. By analyzing
submitted images, the DCNNs use sophisticated pattern
recognition algorithms to detect embedded copyright information
or trademarks, promptly flagging infringements for further action.
The software ensures a seamless user experience with an intuitive
interface that guides users through image upload, copyright
embedding, and infringement detection processes. This
comprehensive approach provides a powerful tool for
safeguarding intellectual property rights in the digital landscape,
offering users an efficient means to protect and enforce copyright
effectively.
KEYWORDS: DCNN, MSB, Copyright, Data Security.
INTELLIGENT SYSTEM FOR DETECTION OF COPYRIGHT-PROTECTED
DATA FOR ENHANCED DATA SECURITY
Ndueso Udoetor1, Godwin Ansa2, Anietie Ekong3, and Anthony Edet4
1Department of Computer Science, Akwa Ibom State University, Mkpat Enin, Nigeria.
Email: Udoetorndueso55@gmail.com
2Department of Computer Science, Akwa Ibom State University, Mkpat Enin, Nigeria.
Email: godwinansa@aksu.edu.ng
3Department of Computer Science, Akwa Ibom State University, Mkpat Enin, Nigeria.
Email: anietieekong@aksu.edu.ng
4Department of Computer Science, Akwa Ibom State University, Mkpat Enin, Nigeria.
Email: anthonyedet73@gmail.com
Cite this article:
Udoetor, N., Ansa, G., Ekong,
A., Edet, A. (2024), Intelligent
System for Detection of
Copyright-Protected Data for
Enhanced Data Security.
British Journal of Computer,
Networking and Information
Technology 7(4), 58-80. DOI:
10.52589/BJCNIT-
OQQNPPCJ
Manuscript History
Received: 11 Aug 2024
Accepted: 6 Oct 2024
Published: 17 Oct 2024
Copyright © 2024 The Author(s).
This is an Open Access article
distributed under the terms of
Creative Commons Attribution-
NonCommercial-NoDerivatives
4.0 International (CC BY-NC-ND
4.0), which permits anyone to
share, use, reproduce and
redistribute in any medium,
provided the original author and
source are credited.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
59 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
INTRODUCTION
In the digital age, the exponential growth of unstructured digital data has presented
unprecedented challenges for protecting intellectual property and preventing copyright
infringement (Basma et al., 2020). As individuals and organizations increasingly rely on digital
platforms to create, share, and store diverse forms of content, the need for robust mechanisms
to safeguard copyrights has become paramount. This research focuses on developing advanced
techniques for the protection and detection of copyrights in unstructured digital data, aiming to
enhance overall data security using a combined effort of Most Significant Bit (MSB) and Deep
learning approaches. Traditional methods of copyright protection often fall short when applied
to unstructured data, which includes images, videos, and textual content with varying formats
and structures. The advent of deep learning, particularly Deep Convolutional Neural Networks
(DCNN), offers a promising avenue for addressing these challenges. DCNNs have
demonstrated remarkable capabilities in feature extraction and pattern recognition, making
them well-suited for analyzing complex and diverse data types. By using the power of DCNNs,
this research seeks to devise a sophisticated copyright protection system capable of identifying
and validating copyrighted material across a wide range of digital formats. Deep learning is
particularly effective in handling large datasets and is well-suited for scenarios involving
multiple classes of data, making it a valuable addition to the proposed framework(Yingqian et
al., 2023). Through the integration of DCNN and Most Significant Bit (MSB) technique, the
research aims to create a comprehensive solution that not only enhances copyright protection
but also contributes to the broader field of data security in the digital world(Li et al., 2017).
The implications of successful implementation of the proposed copyright protection system
extend beyond individual content creators and copyright holders. The safeguarding of
intellectual property in unstructured digital data is crucial for fostering innovation, creativity,
and fair compensation for content producers(Ansa et al., 2012). Additionally, as the digital
space continues to evolve, the research outcomes are expected to contribute significantly to the
ongoing discourse surrounding data security and the responsible use of digital content.
Ultimately, the development of effective and efficient copyright protection mechanisms is
essential for creating a sustainable and secure environment for the digital exchange of ideas
and information(Hoyle et al., 2020). Copyright serves as the legal foundation for protecting the
intellectual property rights of content creators, granting them exclusive rights to reproduce,
distribute, and display their work. In unstructured digital data, which encompasses a vast array
of multimedia content and text, enforcing copyright becomes increasingly complex due to the
sheer volume and diverse nature of digital materials(Shen et al., 2019). Copyright infringement
in this context involves unauthorized use, reproduction, or distribution of copyrighted content,
posing a significant threat to the livelihoods of creators and the integrity of their work. As
technology advances, so do the methods employed by infringers, necessitating innovative
approaches to copyright protection that can adapt to the intricacies of unstructured digital
data(Ekong et al., 2022). The proliferation of digital content and the ease of sharing information
across online platforms have heightened concerns about data security in the context of
copyright protection. Unauthorized access to copyrighted materials not only infringes on the
rights of content creators but also poses risks to the confidentiality and integrity of sensitive
information. The research at hand recognizes the intricate connection between copyright
protection and data security, striving to develop a robust system that not only identifies and
prevents copyright infringement but also fortifies the overall security of digital assets. By
addressing these intertwined challenges, the research aims to contribute to the establishment of
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
60 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
a digital environment where creators can confidently share and disseminate their work without
compromising the security of their intellectual property. In the era of big data, the sheer volume
and diversity of unstructured digital data make traditional copyright enforcement methods
inadequate. As copyright infringement increasingly occurs on a global scale through online
platforms and peer-to-peer networks, the need for sophisticated technologies to detect and
prevent such activities becomes imperative. The proposed use of Deep Convolutional Neural
Networks and Most Significant Bit (MSB) signify a departure from conventional approaches,
offering a more nuanced and adaptive solution to the multifaceted problem of copyright
protection in unstructured digital data. By exploring the intersection of copyright, copyright
infringement, and data security, this research endeavors to provide a comprehensive framework
that not only safeguards the rights of content creators but also contributes to the broader
discourse on responsible digital content dissemination and the protection of intellectual
property in the digital age.
Research Problem
The escalating volume of unstructured digital data in contemporary society has given rise to a
pressing problem concerning the protection of copyrights and the prevention of infringement.
Traditional methods of copyright enforcement prove inadequate in the face of diverse
multimedia content, such as images, videos, and text, which often lack a standardized structure.
As a result, there is an urgent need for advanced technological solutions that can navigate the
complexities of unstructured digital data to identify and mitigate copyright infringement
effectively. This research addresses the overarching problem of safeguarding intellectual
property in an era where the boundaries between legal and illegal digital content usage are
becoming increasingly blurred. The challenge extends beyond the sheer scale of unstructured
digital data; it encompasses the dynamic nature of copyright infringement mechanisms in the
digital space. Rapid advancements in technology have enabled infringers to employ
sophisticated techniques, necessitating a constant evolution in copyright protection methods.
The statement of the problem recognizes the critical gap in existing approaches and emphasizes
the urgency of developing a comprehensive solution that integrates Deep Convolutional Neural
Networks and Steganograph techiniques. The Most Significant Bit (MSB) technique will be
used to protect the data, while deep learning does detection of the protected digital data.
LITERATURE/THEORETICAL UNDERPINNING
This section outlines the theoretical foundation of the research, providing a clear understanding
of the underlying principles and concepts that support the study. It establishes the basis for the
research framework by discussing key theories, models, and prior studies relevant to the topic,
thereby contextualizing the research objectives and methodology. This theoretical groundwork
enables a comprehensive grasp of how the research is built upon established knowledge and
contributes to the existing body of literature.
Copyright Protection in Digital Data
Copyright protection in the area of digital data is an increasingly critical concern in the wake
of the digital revolution (Edet, et al., 2024). As the creation, distribution, and consumption of
content transition to online platforms, safeguarding intellectual property has become
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
61 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
paramount. Copyright protection in digital data refers to the legal and technological measures
implemented to ensure that creators' rights are respected and that unauthorized use or
reproduction of digital content is prevented. This encompasses a wide range of digital assets,
including text, images, audio, video, and software code, among others (Ekong et al., 2024).
Historically, copyright laws have been established to protect creators by granting exclusive
rights to reproduce, distribute, and display their work. However, the advent of the digital era
has posed new challenges as the ease of copying and disseminating digital content has surged.
In response, legal frameworks have evolved, adapting to the nuances of the online environment.
Digital rights management (DRM) technologies have been developed to enforce copyright
protection by restricting access to or usage of digital content through encryption, access
controls, and other mechanisms. These measures aim to strike a balance between the rights of
content creators and the public's need for access to information. Despite these advancements,
copyright protection in digital data faces ongoing challenges(Edet et al., 2024). The borderless
nature of the internet, coupled with the ease of replication and distribution, makes it challenging
to enforce copyright laws globally. Additionally, emerging technologies such as artificial
intelligence and deep learning introduce new complexities, requiring continuous adaptation of
legal and technological frameworks. The ongoing discourse around copyright protection in
digital data underscores the need for a dynamic and collaborative approach involving legal,
technological, and ethical considerations to ensure the sustainable protection of intellectual
property in the digital age. In the digital space, copyright protection plays a crucial role in
fostering creativity and innovation by providing content creators with the assurance that their
intellectual property will be acknowledged and fairly compensated(Ebong et al., 2024). Digital
data, which encompasses a vast array of creative works, faces unique challenges due to its
intangible and easily replicable nature. The ubiquity of online platforms, social media, and file-
sharing services has amplified concerns about unauthorized use, piracy, and the potential
dilution of creators' rights. Copyright protection in digital data involves not only legal
mechanisms but also technological solutions that can adapt to the rapidly evolving digital
environment (Ekong et al., 2024). Digital copyright protection involves the establishment of
clear legal frameworks and international agreements that govern the rights and responsibilities
of content creators, distributors, and consumers. These legal measures aim to strike a balance
between protecting creators' rights and fostering the free exchange of ideas and information.
The Digital Millennium Copyright Act (DMCA) in the United States and similar legislations
worldwide exemplify attempts to adapt copyright laws to the digital era. Additionally,
international organizations like the World Intellectual Property Organization (WIPO) work
towards establishing standardized copyright protection measures on a global scale.
Technologically, copyright protection employs a variety of methods to safeguard digital
data(Uwah & Edet, 2024). Digital watermarks, encryption, and access controls are employed
to embed ownership information and restrict unauthorized access or distribution. Digital Rights
Management (DRM) systems, although contentious, are widely used to manage access to
digital content. However, the effectiveness of these technologies is often debated, with
concerns about their impact on user privacy and the potential for stifling legitimate usage. As
the digital space continues to evolve, copyright protection in digital data remains a dynamic
field that requires ongoing adaptation. Emerging technologies, such as blockchain, are being
explored to create decentralized and tamper-proof systems for tracking and managing digital
rights(Edet & Ansa, 2023). The conversation around copyright protection in digital data
extends beyond legal and technological aspects, involving discussions on ethical
considerations, fair use, and the balance between protecting creators' rights and ensuring access
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
62 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
to information for the broader public. In navigating these complexities, it is essential to foster
a collaborative and multidisciplinary approach that addresses the challenges and opportunities
presented by the digital age.
Related Literature
Shabou et al., 2020, proposed a work on Algorithmic methods to explore the automation of
the appraisal of structured and unstructured digital data. They conducted an interdisciplinary
and innovative research project in Switzerland at the Geneva School of Business
Administration HES-SO, in collaboration with the State Archives of Neuchâtel (Office des
archives de l’État de Neuchâtel, OAEN). The primary objective of the study was to address the
classical challenge of extracting and discriminating relevant data from a vast and diverse range
of data record formats and contents. The focus was on providing a framework and proof of
concept for a software tool aimed at assisting in defensible decision-making regarding the
retention and disposal of records and data at OAEN. The research was structured around two
axes: the archival axis, which involved proposing archival metrics for the appraisal of
structured and unstructured data, and the data mining axis, which aimed to provide algorithmic
methods as complementary or additional metrics for the appraisal process.
In terms of methodology, the exploratory study designed and tested the feasibility of archival
metrics paired with data mining metrics to advance the digital appraisal process systematically
or even automatically. Under Axis 1, the authors undertook three key steps: conceptual
framework design for records data appraisal with a detailed three-dimensional approach
(trustworthiness, exploitability, representativeness); operationalization of proposed metrics in
terms of variables supported by quantitative methods for measurement and scoring; and
validation of the conceptual framework and metrics through feedback from experienced
professionals. This process aimed to assess the relevance and feasibility of the proposed
metrics, demonstrating their acceptability in real-life archival practice. Parallelly, Axis 2
proposed functionalities covering both macro analysis for data and algorithmic methods to
enable the computation of digital archival and data mining. The study thus represents a
comprehensive and practical approach to enhancing the digital appraisal process for archival
and data management purposes.
Haonan et al., 2023 proposed a work on Copyright Protection and Accountability of Generative
AI: Attack, Watermarking and Attribution. This paper focuses on addressing the escalating
concerns surrounding the protection of Intellectual Property Rights (IPR) in the context of
Generative AI, particularly Generative Adversarial Networks (GANs). With the increasing
popularity of Generative AI, the paper highlights the potential risks related to IPR, specifically
in relation to images (toxic images) and models (poisoned models) generated by GANs. The
authors propose an evaluation framework to comprehensively assess the current state of
copyright protection measures for GANs. The evaluation encompasses a diverse range of GAN
architectures and aims to identify factors influencing performance as well as suggest future
research directions. The findings suggest that existing IPR protection methods for input images,
model watermarking, and attribution networks are generally effective for a broad spectrum of
GANs. However, the paper emphasizes the need for increased attention to protecting training
sets, noting that current approaches lack robust IPR protection and provenance tracing for
training data. Overall, the research provides valuable insights into the current state of copyright
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
63 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
protection in the realm of Generative AI, offering directions for improvement and emphasizing
the importance of addressing potential vulnerabilities in training sets.
Cui et al., 2023, proposed a work on diffusion shield: a watermark for data copyright protection
against generative diffusion models. The research presented in this work focuses on addressing
copyright protection concerns arising from the prolific use of Generative Diffusion Models
(GDMs) in generating images. As GDMs gain popularity for their remarkable image generation
capabilities, concerns have emerged regarding the unauthorized replication of creative works
by artists, such as painters and photographers. In response to these challenges, the authors
introduce a novel watermarking scheme called Diffusion Shield. This scheme aims to protect
images from copyright infringement by encoding ownership information into an imperceptible
watermark, which is then injected into the images. The watermark is designed to be easily
learned by GDMs and replicated in their generated images. The proposed method, Diffusion
Shield, utilizes uniform watermarks and a joint optimization approach to ensure low distortion
of the original image, high watermark detection performance, and the capacity to embed
lengthy messages. Rigorous and comprehensive experiments are conducted to demonstrate the
effectiveness of Diffusion Shield in defending against copyright infringement by GDMs,
establishing its superiority over traditional watermarking methods. Overall, the research
provides a practical solution to safeguarding intellectual property in the face of the evolving
capabilities of Generative Diffusion Models.
Liu et al. (2020) concluded several privacy issues of online image sharing. In this paper, the
authors focused on the unawareness of privacy during image sharing. There are two main types
of methods to deal with the risk of unawareness of privacy. The first type of method mainly
adopts classification models to identify private images
Zerr, Siersdorfer, and Hare (2012) proposed a privacy-aware classifer based on vi
sual features like face and color histograms. Buschek et al. (2015) proposed a multi-modal
method that assigns privacy labels to the images based on visual features and metadata like
location and publication time. Tonge, Caragea, and Squicciarini (2018) utilized another kind
of metadata, tag, and Tonge and Caragea (2019) further derived features of the object, scene,
and tags for privacy-leaking image detection. Yang et al. (2020) extracted a knowledge graph
from the images and identifed private images based on object detection and graph neural
networks.
The second type of method focuses on sensitive regions in the images, including approaches
like object detection and semantic segmentation. Some detected private
attributes such as faces (Sun, Wu, and Hoi 2018), license plates (Zhou et al. 2012), and social
relationship (Li et al. 2017a). Orekondy, Schiele, and Fritz (2017) defned a list of privacy
attributes and detected them simultaneously. Some works attempted to protect privacy-leaking
image based on blurring (Fan 2018), blocking (Li et al. 2017b), cartooning (Hasan et al. 2017),
and perturbation (Oh, Fritz, and Schiele 2017). Shetty, Fritz, and Schiele (2018) removed
private objects from the images based on a generative method. However, a person may be
recognized even his face is not visible (Oh et al. 2016), and the redacted image may be
recovered (Shen et al. 2019). As the usage of shared images is almost uncontrollable, it is better
to prevent the risk from the beginning. Therefore, we follow the first type of method to solve
the issue of privacy-leaking images by classification.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
64 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
In exploring Machine Learning capabilities in prediction problems, (Odikwa, Ekong &
Okpako, 2021) demonstrated the use of Fuzzy Logic in the design of Chat-bot Messenger for
Home Intelligent System with Multiple Sensors.Also, (Ekong, James and & Edet, 2022)
proposed a work on Supervised Machine Learning Model For Effective classification Of
Patients With Covid-19 Symptoms-based On Bayesian Belief Network with a focus on the
identification of patients with Covid-19 symptoms. Ekong et al., 2023, used machine learning
algorithm, particularly, the Random Forest algorithm to address a cybersecurity problem. In
another research, Ekong et al., 2023, proposed a work on Machine Learning based Model for
the Prediction of Fasting Blood Sugar Level towards Cardiovascular Disease Control for the
Enhancement of Public Health.In the research, Logistcs Model was adopted for the prediction
of fasting blood sugar level which enabled the control of Cardivascular diseases which in turn
enhanced public health. (Inyang & Umoren, 2023) proposed a work using NLP. This research
employs a Framework-Based Method (FBM) to classify infectious diseases based on ecological
risk factors, providing a structured and reproducible approach. Various machine learning
models are utilized, including XGBoost, Random Forest (RF), Support Vector Machine
(SVM), Artificial Neural Network (ANN), Linear Discriminant Analysis (LDA), Gradient
Boosting Machine (GBM), k-Nearest Neighbors (KNN), and Decision Tree (DT). The results
indicate varying levels of accuracy and Kappa statistics for each model. The study reveals the
performance metrics for each model, with XGBoost and LDA achieving relatively high
accuracy and Kappa values. Additionally, a Deep Learning model, BERT, is integrated with
XGBoost to create an interactive interface for users, enhancing the practical application of
machine learning and Natural Language Processing (NLP) in ecological disease classification.
The research emphasizes the importance of considering ecological risk factors in infectious
disease classification and explores the potential of machine learning techniques for this
purpose. Ekong et al., 2023, proposed a work on Machine Learning Approach for Classification
of SickleCell Anemia in Teenagers Based on Bayesian Network. In this study, a Bayesian
network approach was employed for the classification of sickle cell anemia in teenagers based
on diverse medical data. The research utilized probabilistic graphical models to represent
relationships among various medical parameters, incorporating Bayesian principles for
adaptive predictions. The Bayesian network demonstrated exceptional accuracy (99%) in
classifying teenagers as either positive or negative for sickle cell anemia. Key features
contributing to the classification were identified, offering valuable insights for early detection
and intervention. The findings highlight the diagnostic significance of sickle cell anemia
classification in teenagers, contributing to medical informatics and computational biology. The
Bayesian network proves to be a reliable decision support system for clinicians, aiding in
informed decisions and timely interventions.(Edet and Ansa, 2023), in this research, the authors
proposed a Machine Learning Enabled System for Intelligent Classification of Host-based
Intrusion Severity to effectively manage intrusion events, particularly those initiated by internal
workers. The model comprises three phases: detecting intrusion severity, conducting source
analysis, and providing security recommendations using counterfactual reasoning. The dataset
was gathered from user interactions over time, captured in an activity log. Bayesian Network
achieved an 82% accuracy in the intrusion severity classification. The system includes an API
for scalability, aiming to assist IT firms in analyzing and managing the impact of intrusions
effectively.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
65 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Fig. 1: Architecture of the Existing System (Yingqian et al., 2023)
Yingqian et al. in 2023, proposed a work on diffusion shield: a watermark for data copyright
protection against generative diffusion models. The research presented in this work focuses on
addressing copyright protection concerns arising from the prolific use of Generative Diffusion
Models (GDMs) in generating images. As GDMs gain popularity for their remarkable image
generation capabilities, concerns have emerged regarding the unauthorized replication of
creative works by artists, such as painters and photographers. In response to these challenges,
the authors introduce a novel watermarking scheme called Diffusion Shield. This scheme aims
to protect images from copyright infringement by encoding ownership information into an
imperceptible watermark, which is then injected into the images. The watermark is designed to
be easily learned by GDMs and replicated in their generated images. The proposed method,
Diffusion Shield, utilizes uniform watermarks and a joint optimization approach to ensure low
distortion of the original image, high watermark detection performance, and the capacity to
embed lengthy messages. Rigorous and comprehensive experiments are conducted to
demonstrate the effectiveness of Diffusion Shield in defending against copyright infringement
by GDMs, establishing its superiority over traditional watermarking methods. Overall, the
research provides a practical solution to safeguarding intellectual property in the face of the
evolving capabilities of Generative Diffusion Models. In the existing system, Image + W =
Copyrighted Image. The initial phase of the system involves a crucial computation, taking
place within its first segment. The underlying computational principle employed in this module
is known as the Watermarking principle. This methodology, however, exhibits certain
drawbacks, particularly in comparison to the more advanced machine learning approaches
widely used in the field of data and information security. The Watermarking principle, although
historically employed in the realm of security, is considered less reliable in the contemporary
technological space. This is attributed to a number of weaknesses inherent in watermarking as
an approach to enhancing security systems. Given the rapid evolution of tools designed for
breaching security systems, relying solely on watermarking is no longer deemed sufficient for
constructing robust security frameworks in today's technology-driven environment. The
prevailing recommendation in the development of security systems is the incorporation of
multiple algorithms or the adoption of a single, but exceptionally robust, algorithm. This
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
66 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
strategic combination enhances the overall resilience and effectiveness of the security system.
In contrast, the existing system relies solely on the Watermarking principle, exposing it to
potential vulnerabilities. For the system to remain viable and secure in today's dynamic
technology space, a substantial improvement is imperative. The enhancement should involve
the integration of advanced algorithms, capable of withstanding the sophisticated tools and
techniques employed by malicious actors in attempting to compromise security systems. By
embracing a more comprehensive and sophisticated approach, the system can better adapt to
the evolving landscape of technological threats and ensure its functionality and reliability in
safeguarding sensitive data and information.
Weaknesses of the Existing System
1. Watermarking (Diffusion shield) approach is not robust for copyrighting and protection
of digital images as it relies on obfuscating or altering the image at the pixel level to
protect it.
2. The security level of the existing system is weak and the file is vulnerable or susceptible
to reverse engineering.
3. The process of embedding protective data or watermarks can degrade the quality of the
original image, which might be unacceptable in high-quality or professional applications.
4. Diffusion shield techniques can introduce distortions or artifacts into the image,
potentially affecting its quality or usability. This could be a concern in scenarios where
image fidelity is important.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
67 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Fig. 3: Conceptual Framework of the Proposed System
Input Image to be
Protected
Upload Trade Mark
Calculate MSB of Image to
be Protected
Place Trade Mark in
MSB of Image to be
Protected
IMAGE + W = COPYRIGHTED IMAGE
PUBLIC SPACE
INPUT CONFLICTING
DIGITAL IMAGE
Gather
Metadata
(DCNN)
Image Feature
Analysis
Image
Preprocessing
Protected Not Protected
Show Trademark/Copyright Text
Detect Payload
Storage
External
Users
Usage
Appropriate Authorities
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
68 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
This research focuses on digital image files and the potential risks associated with sharing
unprotected images on the internet. The act of creating and disseminating images without
implementing means of identification not visible to the human eye can jeopardize the rights
to those images. Without proper protection, individuals may lose control over their images,
emphasizing the need for robust security measures in the digital space. Building upon the
work of Yingqian et al. in 2023, this study represents an advancement in image copyright
protection and detection. While the previous work utilized a watermarking approach, a
critical review of existing literature highlights the limitations of watermarking in providing
the necessary security infrastructure for creators and organizations to implement copyright,
safeguard, identify, or detect their digital images on the internet. Consequently, this research
aims to extend and enhance the existing methodology
Fig. 2: Flowchart of the Conceptual System
Trademark
Text
Start
Login
Open MSB
Upload
MSB
Upload
Tradema
Upload
Data
Apply
Decision
Menu
Infring
ement?
Copyright
Text
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
69 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
In this new approach, Deep Learning and Most Significant Bit (MSB) are employed to fortify
digital information by embedding copyright details. This ensures that the original owner can
assert their rights to the image without encountering conflicts. The main system comprises
three distinct modules: the copyright protection module handled by Most Significant Bit
(MSB), copyright detection module handled by machine learning, and file summary module
to further help in the analysis of the image. The file summary module serves as a pivotal point
displaying digital image metadata upon uploading to the main system.
METHODOLOGY
In the proposed system, the Deep Convolutional Neural Network (DCNN) plays a pivotal role
in the development of the copyright protection and detection system. DCNN, renowned for its
effectiveness in image processing and recognition tasks, is harnessed to enhance the system's
capabilities in safeguarding digital images (Ekong et al., 2023). Through its sophisticated
architecture, DCNN is adept at learning hierarchical spatial features from input data, making it
a valuable tool for tasks related to image identification and protection. This approach ensures
that the system benefits from the strengths of DCNN, leveraging the it for image-related tasks
and for effective copyrighted image classification. The integration of deep learning enhances
the system's capacity to detect and protect copyrighted material, thereby providing a well-
rounded solution for digital content security. Most Significant Bit (MSB) is a data security
technique that hides one data in another data allowing it to be hidden from the eyes and
perception of the general public.
1. Deep Convolutional Neural Networks (DCNN)
Deep Convolutional Neural Networks (DCNNs) are a type of artificial neural network designed
for tasks involving visual data, such as image recognition and computer vision. The
mathematical representation of a DCNN involves several components, including convolutional
layers, pooling layers, fully connected layers, and activation functions. Below is a basic outline
of the mathematical operations commonly used in a DCNN:
1.Convolutional Operation:
Input Volume: ( I ) (Image or Feature Map)
Convolutional Filter (Kernel):( K )
Convolution Operation (with Stride ( S )):
[ (I * K)_{i,j} = \sum_m \sum_n I_{(i . S + m),(j . S + n)} K_{m,n}]
Output Feature Map:( O )
[ O_{i,j} = (\sum_m \sum_n I_{(i . S + m),(j . S + n)} K_{m,n} + b)]
where ( f ) is an activation function (e.g., ReLU), and ( b ) is the bias term.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
70 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
2. Pooling Operation:
Max Pooling:
[{MaxPooling}(I)_{i,j} = \max_{m,n} I_{(i . S + m),(j . S + n)}]
Average Pooling:
[ {AvgPooling}(I)_{i,j} = {1}/{m . n} \sum_{m,n} I_{(i . S + m),(j . S + n)}]
3. Fully Connected Layer:
Input Vector: ( X ) (flattened output from previous layers)
Weight Matrix: ( W)
Bias Vector: ( b)
Output Vector:( Y)
[ Y = f(WX + b)]
4. Activation Function:
Commonly used activation functions include ReLU (Rectified Linear Unit):
[ {ReLU}(x) = \max(0, x) ]
and Sigmoid:
[ {Sigmoid}(x) = {1}/{1 + e^{-x}}] ( Ekong et al., 2022)
These mathematical operations are applied across the layers of a DCNN to progressively learn
hierarchical features from the input data (Edet et al., 2024). The network is trained using
backpropagation and optimization algorithms to minimize a defined loss function, ensuring
that the network's output aligns with the ground truth labels for the training data.
2. MSB Algorithm
The Most Significant Bit (MSB) algorithm is a straightforward technique used for data
embedding in digital images. Below is a simple algorithm outlining the steps involved in the
MSB embedding process:
Input:
- Original digital image (represented as a matrix of pixel values)
- Copyright or trademark information to be embedded (binary data)
Output:
- Payload digital image
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
71 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Steps:
1. Convert Copyright Information to Binary:
Convert the copyright or trademark information into binary format. Each character in the
copyright information is represented by its corresponding binary value using ASCII or Unicode
encoding.
2. Iterate Through Image Pixels:
Iterate through each pixel in the digital image, row by row and column by column.
3. Embed Copyright Information:
For each pixel, retrieve the grayscale value or color channels (e.g., RGB values). Since the
MSB is the highest-order bit, it has the least effect on the pixel's intensity or color perception.
Therefore, overwrite the MSB of each pixel's binary representation with the bits of the
copyright information. Repeat this process for each bit of the copyright information until all
bits are embedded.
4. Update Pixel Values:
Update the pixel values in the image matrix with the modified binary representations
containing the embedded copyright information.
5. Output Payload Image:
The resulting image now contains the embedded copyright or trademark information in its
MSB. This image is the output of the MSB embedding algorithm.
3. Strength of MSB and DCNN over the Existing Approach
In this section, we state five potential advantages of using the MSB and DCNN methods over
traditional watermarking techniques for copyright protection in digital images, intellectual
property, trademarks, etc.:
1. Higher Capacity for Data Embedding:
The MSB and DCNN methods can embed more information compared to traditional
watermarking techniques that often use the least significant bits (LSBs). By leveraging the most
significant bits and DCNN, the methods allows for embedding and checking a larger amount
of data without significantly affecting the image quality.
2. Enhanced Robustness:
The MSB and DCNN methods are generally more robust against various types of attacks,
such as compression, noise addition, and other common image processing operations. Since
data is embedded in the more critical parts of the data, it is less likely to be destroyed by such
operations.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
72 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
3. Improved Security:
By embedding data in the most significant bits, the MSB method can offer enhanced security
against unauthorized removal or tampering. Modifying or removing the embedded data without
noticeably altering the image quality can be more challenging for attackers.
4. Better Visual Quality:
The MSB method can maintain better visual quality, especially when the data is spread over
multiple MSBs in a controlled manner. This can result in less noticeable changes compared to
some traditional watermarking methods that may introduce visible distortions.
5. Easier Detection and Verification:
Data embedded using the MSB method can be more easily detected and verified, as it is less
susceptible to loss or degradation during common image processing. This simplifies the process
of verifying copyright and ownership, making it faster and more reliable.
RESULTS/FINDINGS
In this section, the output of the system is presented. Below are the output of the copyright
protection and detection system.
Fig. 3: Application Loading Screen
As the application initiates, it undergoes a meticulous loading process, systematically gathering
essential files and resources vital for its seamless operation. Beginning from index 0 and
progressing sequentially to 100, each file is meticulously retrieved and integrated into the
application's runtime environment. This loading phase ensures that all requisite components,
including libraries, configuration files, and data sets, are efficiently accessed and initialized,
laying a robust foundation for the subsequent execution of the application's functionalities.
During the loading sequence, the application meticulously verifies the integrity and
completeness of each file, performing validation checks to ascertain their authenticity and
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
73 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
compatibility with the runtime environment. This rigorous validation process helps mitigate
potential errors or inconsistencies that may arise during runtime, fostering stability and
reliability in the application's execution. Additionally, the loading phase serves as a critical
preparatory stage, setting the stage for the application to deliver optimal performance and
functionality while adhering to established quality standards and best practices.
Fig. 5: Login Screen
The login screen serves as the gateway to the application's functionalities, ensuring secure
access through robust authentication mechanisms. Upon accessing the login screen, users are
prompted to input their credentials, including a username and password. The system
meticulously validates these credentials against predefined criteria, verifying the correctness
of the provided username and password combination. Only upon successful validation of the
credentials does the login screen grant access to authorized users, enabling them to proceed to
utilize the application's features and functionalities, while maintaining the integrity and
confidentiality of the system.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
74 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Fig. 6: Menu Screen
The menu screen presents users with a pivotal choice, offering two distinct options to tailor
their interaction with the application's functionalities. Users are empowered to select between
protecting and detecting digital images embedded with textual copyright information or those
embedded with trademark images. This clear and intuitive interface design provides users with
flexibility, allowing them to align their actions with their specific copyright protection and
detection needs. Whether opting to safeguard textual content or visual trademarks, users can
seamlessly navigate through the application's features, ensuring comprehensive protection and
detection of their digital assets. By offering these distinct pathways, the menu screen enhances
user engagement and efficiency, enabling users to make informed decisions that align with
their copyright management objectives.
Fig. 7: Screen for Protection of Digital Data with Copyright Text Information
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
75 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
The screen dedicated to the protection of digital data with copyright text information provides
users with a streamlined interface tailored for seamless integration of copyright information
into their digital assets. Upon accessing this screen, users are prompted to upload the image
they intend to safeguard, facilitating a straightforward process for embedding copyright text
for enhanced protection. Additionally, users are prompted to input a password, bolstering
security measures to ensure authorized access to the embedding process. Once the image and
password are provided, the MSB algorithm seamlessly embeds the copyright text into the
digital image, leveraging imperceptible alterations to preserve visual integrity. Following the
embedding process, the image is securely saved as a payload carrying the copyright
information, effectively encapsulating the copyright text within the image's metadata. This
encapsulation ensures that the copyright information remains seamlessly integrated with the
image, serving as a robust identifier in the event of conflict or infringement. By equipping
digital assets with embedded copyright text, users fortify their ability to assert ownership and
defend against unauthorized usage or reproduction. The interface's user-centric design
prioritizes simplicity and efficacy, empowering users to safeguard their intellectual property
with confidence and ease.
Fig. 8: Protection of Digital Data with Trademark Information
The protection of digital data with trademark information follows a meticulous process,
blending the efficiency of the Most Significant Bit (MSB) method with the advanced
capabilities of Deep Convolutional Neural Networks (DCNN) for detection. Users engaging in
this protection process are guided through an interface optimized for seamless integration of
trademark information into their digital assets. With an intuitive design, users are prompted to
upload the image intended for protection, initiating a process where the MSB algorithm
delicately embeds the trademark information into the image's binary data. This embedding
process is conducted with precision, ensuring that the trademark remains discreetly integrated
while maintaining the image's visual fidelity. Upon completion of the embedding process, the
image is fortified with trademark information, serving as a distinct identifier for the protected
asset. During the subsequent detection phase, powered by DCNN, the system rigorously scans
digital content to identify instances of trademark infringement or unauthorized usage.
Leveraging the robust pattern recognition capabilities of DCNN, the system efficiently
identifies trademarked elements within digital assets, providing users with invaluable insights
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
76 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
into potential copyright violations. By integrating the MSB embedding method with DCNN-
based detection, users can effectively safeguard their digital assets, preemptively detecting and
addressing instances of trademark infringement with precision and confidence. This
harmonious fusion of algorithms empowers users to assert ownership rights over their
intellectual property in the digital landscape, ensuring comprehensive protection and
enforcement capabilities.
Fig. 9: Detection of Copyright infringement using DCNN
The Detection of Copyright Infringement using Deep Convolutional Neural Networks (DCNN)
screen provides users with a swift and reliable means to ascertain the copyright status of digital
images, enabling proactive identification of potential infringement or conflicts. Through an
intuitive interface, users can effortlessly upload images for analysis, leveraging the powerful
capabilities of DCNN for copyright detection. Upon submission, the system rigorously
analyzes the uploaded images, employing sophisticated pattern recognition algorithms to
identify copyrighted elements or trademarks embedded within the digital content. This
comprehensive analysis enables users to swiftly determine the copyright status of images,
empowering them to take appropriate action in cases of infringement or conflict. With its user-
friendly design and efficient processing capabilities, the DCNN-based detection screen serves
as a valuable tool for safeguarding intellectual property rights in the digital domain, facilitating
proactive enforcement and protection measures with ease.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
77 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Fig. 10: Program files from folder
The program files necessary for system programming are meticulously organized within
designated folders, ensuring efficient development and maintenance workflows. These folders
house a comprehensive array of files essential for the system's functionality, including source
code, libraries, configuration files, and documentation.
IMPLICATION TO RESEARCH AND PRACTICE
The implications of this research to practice are significant, as it offers a practical framework
for enhancing copyright protection and detection in the digital domain. By integrating advanced
algorithms like the Most Significant Bit (MSB) embedding technique and Deep Convolutional
Neural Networks (DCNNs), the software provides a robust solution for safeguarding
intellectual property rights against unauthorized use. Practitioners can utilize this technology
to efficiently embed copyright information into digital images and detect potential
infringements, thus streamlining enforcement processes and reducing the risk of intellectual
property theft. This practical application empowers individuals and organizations to better
protect their digital assets and uphold their rights, contributing to a more secure and equitable
digital nvironment.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
78 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
CONCLUSION
In conclusion, this research has developed and evaluated a software solution designed to
address the growing concerns of copyright infringement in the digital landscape. The software
employs a dual-faceted approach: during the protection phase, it utilizes the Most Significant
Bit (MSB) embedding technique to subtly integrate copyright or trademark information into
digital images, ensuring that this data is not perceptible while maintaining the image's visual
quality. During the detection phase, the software leverages Deep Convolutional Neural
Networks (DCNNs) to rigorously analyze images for signs of unauthorized use or copyright
violations. This methodology combines advanced image processing with sophisticated pattern
recognition to offer a comprehensive tool for protecting and enforcing intellectual property
rights. The results of this research demonstrate the effectiveness of the proposed solution in
both embedding copyright information and detecting infringement. The MSB embedding
technique proved to be efficient in encoding protection data without compromising image
integrity, while the DCNN-based detection algorithm successfully identified instances of
copyright violations with high accuracy. This combination of methodologies not only enhances
the capability to safeguard digital assets but also streamlines the enforcement process for users.
The software’s intuitive interface ensures that users can navigate these functionalities with
ease, making it a valuable asset for individuals and organizations seeking to protect their
intellectual property in an increasingly digital world.
FUTURE RESEARCH
Future research should explore integrating additional machine learning models and encryption
techniques to further enhance the robustness and adaptability of copyright protection and
detection mechanisms .
REFERENCES
Anietie Ekong, Blessing Ekong and Anthony Edet (2022), Supervised Machine Learning
Model for EffectiveClassification of Patients with Covid-19 Symptoms Based on
Bayesian Belief Network, Researchers Journal ofScience and Technology(2022),2, pp-
27-33.
Ansa, G., Cruickshank, H., & Sun, Z. (2012). An Energy-Efficient Technique to Combat DOS
Attacks in Delay Tolerant Networks. EAI Endorsed Transactions on Ubiquitous
Environments, 1, (1),.
Basma, M. S., Julien, Tièche, J. K. & Arnaud, G. (2020). Algorithmic methods to explore the
automation of the appraisal of structured and unstructured digital data, Records
Management Journal,30(2), 175-200.
Buschek, D., Bader, M., von Zezschwitz, E., & Luca, A. (2015). Automatic privacy
classification of personal photos. In IFIP Conference on Human-Computer Interaction,
428–435.
Ebong, O., Edet, A., Uwah, A., & Udoetor, N. (2024). Comprehensive Impact Assessment of
Intrusion Detection and Mitigation Strategies Using Support Vector Machine
Classification. Research Journal of Pure Science and Technology, 7,(2), 50-69.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
79 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Edet, A. E. and Ansa, G. O. (2023). Machine learning enabled system for intelligent
classification of host-based intrusion severity. Global Journal of Engineering and
Technology Advances,16(03), 041–050.
Edet, A., Ekong, B. and Attih, I. (2024). Machine Learning Enabled System for Health Impact
Assessment of Soft Drink Consumption Using Ensemble Learning Technique.
International Journal Of Computer Science And Mathematical Theory,10(1):79-101,
DOI: 10.56201/ijcsmt.v10.no1.2024.pg79.101
Edet, A., Silas, A., Ekaetor, E., Ebong, O., Isaac, E., & Udoetor, N. (2024). DATA-DRIVEN
FRAMEWORK FOR CLASSIFICATION AND MANAGEMENT OF START-UP
RISK FOR HIGH INVESTMENT RETURNS. Advanced Journal of Science,
Technology and Engineering, 4,(2),81-102.
Edet, A., Udonna, U., Attih, I., and Uwah, A. (2024). Security Framework for Detection of
Denial of Service (DoS) Attack on Virtual Private Networks for Efficient Data
Transmission. Research Journal of Pure Science and Technology, 7(1),71-81. DOI:
10.56201/rjpst.v7.no1.2024.pg71.81
Ekong, A., James, G., Ekpe, G., Edet, A., & Dominic, E. A (2024). Model For The
Classification Of Bladder State Based On Bayesian Network. International Journal of
Engineering and Artificial Intelligence, 5 ,(2) 33–47
Ekong, B., Edet, A., Udonna, U., Uwah, A.,and Udoetor, N. (2024), Machine Learning Model
for Adverse Drug Reaction Detection Based on Naive Bayes and XGBoost Algorithm.
British Journal of Computer, Networking and Information Technology 7(2), 97-114.DOI:
10.52589/BJCNIT-35MFFBC6
Ekong, B., Ekong, O., Silas, A., Edet, A., & William, B. (2023). Machine Learning Approach
for Classification of Sickle Cell Anemia in Teenagers Based on Bayesian Network.
Journal of Information Systems and Informatics, 5(4), 1793-1808.
https://doi.org/10.51519/journalisi.v5i4.629.
Haonan, Z. , Jiamin, C., Ziyue, Y., Tingmin, W., Pathum, C. M. A. , Chehara, P., & Minhui,
X. (2023). Copyright Protection and Accountability of Generative AI: Attack,
Watermarking and Attribution. University of New South Wales, Australia.
Hasan, R., Shaffer, P., Crandall, D., Apu Kapadia, E. T., et al. (2017). Cartooning for enhanced
privacy in lifelogging and streaming videos. In Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition Workshops, 29–38.
Henry, O., Ekong, A., Okpako, E. (2021). A Fuzzy-Based Chat-bot Messenger for Home
Intelligent System with Multiple Sensors. International Journal of Advanced Trends in
Computer Science and Engineering,10,(2),1026-1038.
Hoyle, R., Stark, L., Ismail, Q., Crandall, D., Kapadia, A., and Anthony, D. (2020). Privacy
Norms and Preferences for Photos Posted Online. ACM Transactions on Computer
Human Interaction, 27(4): 1–27.
Li, J., Wong, Y., Zhao, Q., and Kankanhalli, M. S. (2017a). Dual-glance model for deciphering
social relationships. In Proceedings of the IEEE International Conference on Computer
Vision, 2650–2659.
Liu, C., Zhu, T., Zhang, J., and Zhou, W. (2020). Privacy Intelligence: A Survey on Image
Sharing on Online Social Networks. arXiv preprint arXiv:2008.12199.
Oh, S. J., Benenson, R., Fritz, M., & Schiele, B. (2016). Faceless person recognition: Privacy
implications in social media. In European Conference on Computer Vision, 19– 35.
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 7, Issue 4, 2024 (pp. 58-80)
80 Article DOI: 10.52589/BJCNIT-OQQNPPCJ
DOI URL: https://doi.org/10.52589/BJCNIT-OQQNPPCJ
www.abjournals.org
Oh, S. J., Fritz, M.,& Schiele, B. (2017). Adversarial image perturbation for privacy protection
a game theory perspective. In Proceedings of the IEEE International Conference on
Computer Vision, 1491–1500.
Orekondy, T., Schiele, B., and Fritz, M. (2017). Towards a visual privacy advisor:
Understanding and predicting privacy risks in images. In Proceedings of the IEEE
International Conference on Computer Vision, 3686–3695.
Shabou, B., Tièche, J., Knafou, J., Gaudinat, A. (2020). Algorithmic methods to explore the
automation of the appraisal of structured and unstructured digital data. Records
Management Journal. DOI: 10.1108/RMJ-09-2019-0049
Shen, L., Hong, R., Zhang, H., Zhang, H., and Wang, M. (2019). Single-shot Semantic Image
Inpainting with Densely Connected Generative Networks. In Proceedings of the 27th
ACM International Conference on Multimedia, 1861–1869.
Shetty, R. R., Fritz, M., and Schiele, B. (2018). Adversarial scene editing: Automatic object
removal from weak supervision. In Advances in Neural Information Processing Systems,
7706–7716.
Sun, X., Wu, P., and Hoi, S. C. (2018). Face detection using deep learning: An improved faster
RCNN approach. Neurocomputing, 299: 42–50.
Tonge, A., and Caragea, C. (2019). Dynamic deep multimodal fusion for image privacy
prediction. In The World Wide Web Conference, 1829–1840.
Tonge, A., Caragea, C., and Squicciarini, A. (2018). Uncovering scene context for predicting
privacy of online shared images. In Proceedings of the 32nd AAAI Conference on
Artifcial Intelligence, 8167–8168.
Uwah, A. and Edet, A. (2024).Customized Web Application for Addressing Language
ModelMisalignment through Reinforcement Learning from HumanFeedback. World
Journal of Innovation And Modern Technology,8,(1), 62-71. DOI:
10.56201/wjimt.v8.no1.2024.pg62.71.
Yang, G., Cao, G., Chen, Z., Guo, J., & Li, J. (2020). Graph-based neural networks for
explainable image privacy inference, Pattern Recognition, Volume 105,
https://doi.org/10.1016/j.patcog.2020.107360.
Yingqian, C., Jie, R., Han, X., Pengfei, H., Hui, L., Lichao, S., Yue, X., Jiliang, T. (2023).
Diffusionshield: a Watermark For Data CopyRight Protection Against Generative
Diffusion Models.
Yingqian, C., Jie, R., Han, X., Pengfei, H., Hui, L., Lichao, S., Yue, X., Jiliang, T. (2023).
Diffusionshield: a Watermark For Data CopyRight Protection Against Generative
Diffusion Models.
Zerr, S., Siersdorfer, S., Hare, J., and Demidova, E. (2012). Privacy-aware image classifcation
and search. In Proceedings of the 35th International ACM SIGIR Conference on Research
and Development in Information Retrieval, 35–44.
Zhou, W., Li, H., Lu, Y., and Tian, Q. (2012). Principal visual word discovery for automatic
license plate detection. IEEE Transactions on Image Processing, 21(9): 4269–4279.
