Pak. j. life soc. Sci.

(2024), 22(2): 9519-9535 E-ISSN: 2221-7630;P-ISSN: 1727-4915

Pakistan Journal of Life and Social Sciences
www.pjlss.edu.pk

https://doi.org/10.57239/PJLSS-2024-22.2.00721

RESEARCH ARTICLE

The Future of Criminal Law in the Age of Electronic Crimes and

Artificial Intelligence
Dr. Ahmad Hazim Mustafa
Salahaddin University, College of Law, Erbil, Iraq

ARTICLE INFO ABSTRACT

As technological advancements continue to reshape our world, the realm of
Received: Aug 16, 2024
criminal law faces unprecedented challenges and opportunities. The
Accepted: Oct 22, 2024 proliferation of electronic crimes and the rise of artificial intelligence (AI)
have generated complex legal, ethical, and practical considerations for
criminal justice systems worldwide. This research paper delves into the
Keywords multifaceted impact of electronic crimes and AI on criminal law, examining
how they alter traditional legal paradigms, discussing emerging legal
Electronic Crimes
frameworks, and addressing the implications for individual rights and
Criminal Law societal norms. By exploring various case studies and analyzing the
evolution of legislation, this paper seeks to provide insight into the evolving
Artificial Intelligence landscape of criminal law in the age of electronic crimes and artificial
intelligence
*Corresponding Author:

INTRODUCTION
In the contemporary context of the swiftly progressing digital environment, the issue of information
security has arisen as a matter of utmost importance. The advent of the internet and networked gadgets
has brought us a period of unparalleled ease, however it has also exposed consumers to the escalating
dangers of electronic crimes. The emergence of the virtual realm has led to the emergence of a novel
category of weaponry known as digital assaults. These attacks only occur within the domain of
cyberspace, yet possess a level of effectiveness that is comparable to conventional physical threats.
Among the various types of cyberattacks, those directed on critical infrastructure, such as power grids
and military systems, present notably grave risks, since they have the capacity to cause catastrophic
failures.
The potential ramifications of these technological offenses can be significant. Cyber attacks has the
capability to disrupt computer networks and telecommunications systems, resulting in the
inaccessibility of crucial data and jeopardizing the performance of essential services. As the dependence
of society on networked technology grows, there is a growing imperative to address the associated
hazards.
Significantly, financial and medical organizations have emerged as primary targets for these assaults
because to the sensitive nature of the data they manage. The rapid increase of frameworks, data
repositories, cloud services, apps, and devices has significantly enlarged the attack surface, hence

9519
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

providing bad actors with a greater number of possibilities to exploit weaknesses inside the network.
In light of this, businesses and organizations are endeavoring to attain a position of superiority in this
digital realm by using the capabilities of artificial intelligence (AI) and other emerging technologies.
The utilization of artificial intelligence, characterized by its capacity to do accurate mathematical
calculations and complex numerical evaluations, has demonstrated its efficacy as a powerful asset in
safeguarding the authenticity and security of the digital domain. Significant applicability has been
observed in the detection and mitigation of unforeseen digital assaults across various cyberspaces,
encompassing web-based platforms as well as highly fortified websites. One of the main functions of
this system is to detect illicit content on government websites and identify cyberattacks in their early
phases, enabling the implementation of preventive countermeasures.
Artificial intelligence (AI) serves as a pivotal component in ensuring online security, therefore
instigating a significant transformation in the domain of criminal law. The utilization of computational
reasoning techniques presents novel opportunities for enhancing cybersecurity measures in response
to the increasing challenges faced by electronic criminal activities. Artificial intelligence (AI) systems
has the capability to rapidly analyze extensive quantities of data, enabling them to identify patterns that
may signify hostile activity. Consequently, these systems can promptly react to counteract such threats.
Nevertheless, the integration of artificial intelligence (AI) inside the criminal law framework presents
a distinct array of obstacles. The legal framework should be modified to effectively address the
complexities of electronic offenses and the dynamic strategies adopted by those engaging in
cybercriminal activities. Achieving an optimal equilibrium between upholding individual privacy rights
and facilitating the implementation of efficient AI-based security measures is a nuanced undertaking
that needs meticulous deliberation.
Problem Statement
The subject of criminal law has been confronted with new issues and complications due to the fast
progress of technology, namely in the areas of electronic communication and artificial intelligence. The
increasing interconnectivity and dependence on digital systems inside society have led to a significant
rise in electronic crimes, necessitating the adaptation and evolution of legal systems. In addition, the
incorporation of artificial intelligence (AI) into many aspects of everyday existence gives rise to
significant inquiries on legal responsibility, privacy, and the limits of criminal culpability.
The objective of this issue statement is to tackle the complex difficulties presented by electronic crimes
and artificial intelligence within the framework of criminal law. The primary aims of this study are to
comprehend, evaluate, and provide resolutions for the subsequent significant concerns:
The present discourse aims to delineate and classify electronic crimes, encompassing activities such as
hacking, identity theft, online fraud, and cyberbullying. It is evident that the exponential growth of these
offenses has beyond the capacity of the legal framework to effectively address them. The precise
definition and categorization of these offences are of utmost importance in ensuring efficient law
enforcement and equitable judicial proceedings.
The issue of jurisdiction and international cooperation is particularly salient in the context of electronic
crimes, since these offenses frequently transcend national boundaries, hence presenting intricate
issues in terms of determining legal authority. The establishment of mechanisms for international
collaboration in the investigation and prosecution of these crimes is imperative in order to guarantee
the accountability of offenders, irrespective of their geographic whereabouts.
The incorporation of artificial intelligence (AI) across several domains, such as law enforcement, gives
rise to concerns regarding the legal and ethical ramifications of choices made by AI systems. The
difficulty of ascertaining the allocation of legal culpability in circumstances involving acts or choices
created by artificial intelligence is of considerable importance.

9520
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

The utilization of electronic surveillance techniques and artificial intelligence algorithms in the context
of predictive policing raises significant concerns regarding the potential violation of individual privacy
rights. The delicate task of reconciling the need of safeguarding public safety with the imperative of
upholding civil freedoms necessitates meticulous legal analysis.
The problem lies in guaranteeing the admissibility and accuracy of digital evidence during judicial
proceedings. The necessity for updated evidence rules is emphasized by the technical challenges related
to data integrity and the possibility for manipulation.
In order to successfully investigate, prosecute, and defend cases in the ever-changing environment of
electronic crimes and artificial intelligence (AI), it is imperative for legal professionals and law
enforcement organizations to possess a comprehensive understanding of these complex domains. This
necessitates the development of capacity building initiatives and the acquisition of legal expertise in
the nuances of electronic crimes and AI. It is crucial to address the knowledge disparity.
The regulation of artificial intelligence (AI) in the context of criminal activities is a pressing concern due
to its potential exploitation for illicit purposes, including the creation of malicious software and the
execution of sophisticated fraudulent schemes. It is of utmost importance to establish policies that
effectively tackle these developing concerns, all the while promoting the responsible development of
artificial intelligence.
The determination of suitable penalties for electronic crimes, taking into account the intangible
character of harm, necessitates the use of creative methodologies. Furthermore, it is imperative that
the rehabilitation of criminals is in accordance with the dynamic and evolving technology environment.
The objective of this issue statement is to promote collaborative study across several disciplines,
including legal academics, technologists, ethicists, policymakers, and law enforcement organizations.
The objective is to formulate inclusive and flexible approaches within the context of criminal law
frameworks in order to effectively tackle the difficulties presented by electronic crimes and the
emergence of artificial intelligence. In an era characterized by the proliferation of digital technologies
and the rise of artificial intelligence, it is imperative to devise efficacious strategies that prioritize the
preservation of justice, the protection of civil rights, and the preservation of the rule of law..
Research Questions
Examining the Dynamic Terrain of Criminal Law: Navigating the Realm of Electronic Offenses and
Artificial Intelligence. The present study aims to examine the profound effects of electronic crimes and
artificial intelligence on the domain of criminal law. This study seeks to predict the necessary legal
adjustments to successfully manage new dangers posed by cybercrimes, including hacking and identity
theft, as well as the increasing involvement of artificial intelligence in criminal activities. Through an
analysis of these difficulties, the study attempts to provide insights into the appropriate legal measures
needed to tackle these evolving issues. This analysis explores many aspects related to jurisdiction,
privacy, evidential standards, and responsibility within the dynamic framework under consideration.
This research makes a valuable contribution to the field by examining the complex relationship
between technology improvements and legal frameworks. By analyzing this interaction, the study aims
to enhance the creation of legal tactics that are both adaptable and successful. The ultimate goal is to
safeguard justice and security in the midst of evolving criminal landscapes.
Relevance and Significance
In contemporary society, a wide range of activities, spanning from individuals' personal lives to their
financial transactions, are being handled through online platforms. All elements inside our environment
are interconnected, accompanied by various forms of adaptable processing of electronic data. The
numerous advancements that have been made have resulted in an increased susceptibility and
vulnerability of human life to potential damage. This phenomenon possesses the capacity to engender

9521
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

financial calamities, individuals engaging in dishonest practices, and other unfavorable outcomes.
Therefore, it is imperative to prioritize the safeguarding of networks in the contemporary era of
extensive automation. It offers individuals with a sense of security, safeguarding them against
fraudulent activities and many types of violations. Nevertheless, the efficacy of digital protection will
be subject to scrutiny. Both individual individuals and major organizations alike are increasingly
susceptible to various cyber threats, including phishing scams, ransomware attacks, personal assaults,
data breaches, and financial damages. According to a NetScout investigation, it has been observed that
malevolent actors possess the capability to infiltrate any internet-connected gadget, including but not
limited to smartphones, wristwatches, computers, and televisions, within a very short span of five
minutes.
Currently, the predominant function of artificial intelligence (AI) is to deploy sophisticated, knowledge-
based systems capable of efficiently and precisely managing a wide range of cybercrimes, both of little
and big significance. The advancement of artificial intelligence necessitates the capability of AI
applications to effectively respond and adjust to emerging obstacles and dangers. The utilization of
artificial intelligence is imperative in network security due to the substantial magnitude of sensitive
data being communicated across the internet. In order to effectively handle the vast quantities of data
and mitigate the risk of unauthorized access, the implementation of a network security framework
reliant on artificial intelligence is needed. The utilization of this tool has the potential to enhance the
efficiency of problem identification. It is quite probable that cybercriminals have already gained
unauthorized access to these systems and are already biding their time for an opportune moment to
execute their malicious activities. Artificial intelligence possesses the capacity to effectively perform
rapid situational analysis, recognize potential threats, and implement appropriate measures for
mitigation. Nevertheless, it has become a conventional instrument for achieving the highest level of
network security and efficiently handling substantial volumes of data to enhance threat detection and
optimize response strategies.
LITERATURE REVIEW
The methodologies employed and the objectives pursued in the many crime prediction research and
applications documented in our collected publications exhibited considerable variation. Numerous
crime prediction methodologies have been devised for general crimes and scenarios, employing diverse
models that have been evaluated in order to identify the most efficacious approach based on the given
dataset. In a study done by Kim et al. (2019a), machine learning techniques were employed to forecast
general criminal activities. Nevertheless, several approaches have been devised to address certain
types or classifications of crimes. For instance, Srivastava et al. (2008) employed a hidden Markov
model (HMM) to analyze the sequence of activities in credit card transactions. Previous studies have
concentrated on conducting a comparative examination of various learning model types. For instance,
Babakura et al. (2015) conducted a study in which they compared two classification algorithms, namely
naïve Bayes and backpropagation, to predict crime categories using a specific dataset. The experiment
was conducted using a 10-fold cross-validation technique. The results indicate that the naïve Bayes
algorithm outperformed the backpropagation algorithm when applied to the crime dataset using the
Weka software. Furthermore, a subset of the publications exhibited notable distinctions with respect
to their stated goals. For instance, Nasridinov et al. (2013) conducted a study whereby they utilized
several learning models and algorithms to examine their efficacy across diverse datasets. The
researchers reached the conclusion that the selection of a model type should be done with careful
consideration of the dataset at hand, since different model types exhibit varying degrees of
compatibility with certain datasets.

9522
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Data Mining Techniques for Crime Prediction

Conversely, several recent surveys have investigated techniques for crime prediction. Saravanan et al.
(2021) conducted a survey that examines several data mining techniques for crime prediction, focusing
on elements such as socioeconomic indicators, spatial-temporal patterns, demographic characteristics,
and geographic qualities. Moreover, Butt et al. (2020) conducted a comprehensive assessment of the
literature, focusing on the identification and forecasting of spatiotemporal crime hotspots. The authors
presented their findings and insights in this regard. The authors presented the methodologies of
machine learning (ML) and data mining in the context of hotspot identification. They also discussed the
efficacy of these techniques and highlighted the difficulties associated with constructing a
spatiotemporal crime prediction model. In a separate study, Kawthalkar et al. (2020) conducted a
comprehensive analysis of several technology mapping approaches used in the context of crime
prediction inside smart cities. The researchers considered many variations of criminal portrayals and
conducted a comparison analysis. The authors posit that a multitude of concepts and methodologies
have been established for the purpose of crime prediction; nonetheless, they contend that the
practicality and effectiveness of these techniques can only be ascertained via rigorous field testing.
Furthermore, the authors in the study conducted by Albo (n.d.) place their emphasis on artificial neural
networks and convolutional network approaches in the context of crime prediction.
Urban Safety and Security in Crime Analysis
Safety and security are crucial elements that enhance the overall quality of life in urban environments.
In their work, Zhao and Tang (2018) provided a comprehensive review that encapsulated the field of
crime analysis in urban data. They examined several criminal task algorithms and delved into the
theoretical underpinnings of criminology. Furthermore, Shamsuddin et al. (2017) conducted a concise
and accessible survey about the utilization of crime prediction tools and the potential for their
enhancement in further research. The researchers employed Support Vector Machines (SVM), fuzzy
theory, artificial neural networks, and multivariate time series as machine learning techniques. In
contrast, a comprehensive analysis and prediction of criminal activities were conducted by Fredrick
David KR et al. (2017), who examined both supervised and unsupervised methodologies for crime
detection
APPROACH AND METHODOLOGY
The fundamental principles of information security theory as they pertain to the management of
extensive repositories of data. The use of this resides in the application of artificial intelligence and its
diverse range of methodologies. These tactics can be subject to various alterations, and there is an
ongoing development of new resources to accommodate their evolving usage and scope. These
measures can be included into pre-existing or newly developed software to mitigate or eradicate cyber
security risks and attacks. The use of this technology extends to several practical domains, such as email
security, detection of phishing attacks, identification of malware, detection of anomalies in network
behavior, prevention of network attacks, safeguarding sensitive data, prevention of fraudulent
activities, and mitigation of identity theft risks. The future talks will focus on the idea of applying
various algorithms for the identification of strategies.
Using Artificial Intelligence to Recognize Potential Dangers in Email
Due to technological developments, the quantity of email traffic is increasing. Currently, email is the
most secure mode of internal business communication for confidential information. This has led to an
increase in attempts to prevent spam emails from being sent from corporations and other institutions.
Email is the most common vector for cyberattacks because of the sheer volume of information it
transports. The creation of algorithmic methods of machine learning is urgently required. The
perceptron, in its purest form, is analogous to a neuron in the brain. By constructing layered structures
at varying input levels, the perceptron model can achieve a functioning that is functionally comparable.

9523
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

The input data is turned into the output based on the importance of the synthesized input values. These
weighted input values are then used to trigger an activation function at a predetermined threshold. The
key distinction between classical models and AI models is the latter's usage of iterations to attain
optimal value. To make the most of AI for spam detection, it is necessary to create tasks at which spam
filters can excel. The filters' algorithms are in charge of sorting emails based on whether or not they
include a set of potentially malicious phrases. The frequency with which certain suspicious words or
symbols appear can be used to assign a weight to them. Once the cutoff point is established, the
incoming emails may be sorted accordingly. By definition, ham messages have values below the
threshold limit, whereas spam messages have values over the limit. The AI is supposed to use the data
about spam and ham to fine-tune the threshold value.
Initial spam detections are created using static rules and regular expressions. A threshold study must
be performed, and new approaches must be developed, to bring the spam filters up to date. This is
because spammers are constantly adapting their strategies. The filters need to be constantly updated,
thus it's best to choose a dynamic approach, since static rules may become outdated very soon. The
user's participation is also required for success in this endeavor. Numerous methods exist for verifying
text, including computer vision and natural language processing, which may be used to spot any
discrepancies or questionable passages.
Malware Analysis and DetectionMalware, short for "malicious software," is created by hackers and
other bad actors with the goal of infecting a single computer or a network of computers. The objective
is to get unauthorized access to the organization's data by means of an assault on the system or
networks or by exploiting vulnerabilities in such systems or networks. If the assault is successful, it
might lead to the loss of private data and possible harm to the network infrastructure. Threats of this
nature pose the greatest risk to the system. Spam emails are sent with the intent of spreading malware
across a computer network. It is crucial to detect any malicious software since it is simple to conduct
additional assaults after a machine or network has been breached. As a result, identifying the binary
files that make up the malware is the first step in the analysis process. These binary files must be
partitioned from any files that might contain malicious code before being saved to the system.
Sometimes even non-executable file types like documents might harbor malware. An infected internet
connection, generally a local area network or wireless network, compromises the machine files on an
infected computer.Human intervention is no longer practical for reliably recognizing potentially
harmful threats in light of the ever-increasing influx of information. To counteract this, experts have
created algorithms to analyze malware automatically. In this method, a professional in malware
conducts an initial investigation. After that, AI technologies are used in these assessments to classify
the numerous dangers that have been identified. Analysts often adjust analyses in order to enhance
algorithm updates and hence the defenses against cyberattacks. The procedure of fine tuning can be
done either automatically or by hand. Malware poses a wide range of potential threats. A handful of
them will be discussed in the following paragraphs:
Trojans are a type of harmful software that have the ability to disguise themselves as legitimate
applications. The major purpose of this particular form of malicious software is to gain unauthorized
access to the user's computer system and assume control over it.
Downloaders are a type of software that, when launched, can result in the installation of malware on a
computer. Once an internet connection is established, this malware allows infected distant servers to
gain unauthorized access to the compromised system.
Rootkits are a type of files that are specifically crafted to evade detection inside the operating system
of a user, all the while acquiring unfettered privileges to access the user's data.
Botnets are a form of malicious software that is orchestrated by individuals with malicious intent or
criminal motives, with the objective of acquiring personal or financial information.

9524
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

A zero-day exploit refers to a type of malicious software that lacks antivirus signatures. The lack of
these fingerprints hinders the system's ability to discern the pathogen.
Ransomware is a type of malicious software that involves the infiltration of a computer system with the
intention of preventing the user from accessing specific critical documents or data. Access to the
premises is contingent upon the completion of a monetary transaction; thus, it is imperative that you
promptly proceed with the payment.
These many circumstances can be merged into a unified file, with the subsequent course of action being
determined by the system's security settings administration. Given the extensive range of malware in
existence, it is imperative to devise distinct approaches for the identification of each specific threat.
Calculating file hashes, monitoring system and network activities, and doing system monitoring are all
crucial undertakings. The identification of potential hazards inside the system can be facilitated by the
utilization of hash file computations. System monitoring becomes important when either the hardware
or the software exhibits abnormal behavior. The necessity of network monitoring arises when there is
evidence of atypical associations, whether observed on the internet or inside the internal network of
an organization.
Attacks from the Internet and Strange Behavior from Networks
In recent years, there has been a significant increase in the prevalence of interconnected devices. The
obsolescence of traditional approaches to perimeter security has been brought about by the increasing
prevalence of network-based breaches. Therefore, it is imperative to utilize automated methods for
detecting possible security breaches within the network. One such strategy that may be employed is
the utilization of the signature-based detection method. This approach may be utilized to establish a
comprehensive repository of attack signatures that have been previously detected. The integration of
this database with others results in the activation of an alarm system if the presence of questionable
signatures is detected. The artificial intelligence has the capability to automate the task of ensuring the
database remains up-to-date, a crucial aspect in this particular context. The subsequent stage is
identifying any inconsistencies. This technique involves the monitoring and analysis of network activity
in order to create a baseline for typical behavior. Data is collected and assessed about the typical
patterns exhibited by traffic, encompassing the quantity of connections coming from a certain host, any
atypical connections, any instances of heightened traffic, and any fluctuations in network capacity. An
anomaly may be characterized as any departure from the established norm or a potentially
questionable trend observed within the dataset.
FINDINGS AND RESULTS
Methods for Crime Prediction
Supervised learning and unsupervised learning are the primary methodologies employed in the fields
of artificial intelligence (AI) and machine learning (ML), respectively. The primary differentiation
between supervised and unsupervised learning is in the utilization of labeled data for direct prediction.
There are other variations, exclusions, and significant domains in which one technique demonstrates
superiority over the other.
The utilization of labeled datasets is of paramount importance in the implementation of the supervised
learning approach. The purpose of these data sets is to be utilized in the training and monitoring of
algorithms for accurate data identification and prediction. By utilizing the annotated inputs and
corresponding outputs, the model will autonomously assess its own correctness and progressively
enhance its performance over a period of time. Moreover, within the realm of supervised learning, there
are two distinct categories of problems, namely classification and regression.
In order to effectively differentiate between dogs and cats, the field of Machine Learning employs a
classification algorithm to classify and categorize test data. Another practical example is the act of

9525
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

filtering one's email inbox to identify and subsequently delete unsolicited and unwanted
communications, commonly referred to as spam.
The differentiation between supervised learning and unsupervised learning lies in the presence of
labeled input and output data in the former, whereas the latter lacks such labeling. In the context of
supervised learning, the algorithm is provided with feedback in the form of corrected predictions and
weights, which are derived from the training data. The objective of unsupervised learning is to train a
model to comprehend the inherent structure of unlabeled data. In contrast, semisupervised learning is
a hybrid approach that can be employed when the training dataset comprises both labeled and
unlabeled instances. The performance of the semisupervised technique is enhanced when the number
of labeled samples in the dataset is smaller than the number of unlabeled samples. Despite its
reputation for complexity, researchers continue to utilize this strategy. In the study conducted by
Tundis et al. (2019), the authors employed a semi-supervised approach utilizing bags of words to assess
textual data, namely tweets pertaining to criminal activities, in the context of A108.
According to the depicted figure, around 31% of the existing research on crime prediction pertains to
the domain of supervised learning. Furthermore, it is worth noting that 22% of the overall research
articles used a combination of supervised and unsupervised methods. This is attributed to the fact that
several studies utilized more than one machine learning methodology. However, just a mere 10% of
individuals employed unsupervised learning techniques. Remarkably, a mere 1% of the investigations
implemented a semisupervised approach, suggesting that the utilization of this technique in the realm
of crime prediction is infrequent. Ultimately, a notable proportion of the research, specifically 36%,
exhibited a lack of clarity on the specific approach employed.

Benefits and limitations associated with contemporary prediction technologies.

The random forest model has gained significant popularity in the field of crime prediction due to its
ability to effectively handle colinear information, high dimensionality, and heterogeneous space (Rumi
et al., 2018; Umair et al., 2020; Wheeler & Steenbeek, 2020). Additionally, A93 has discovered that the
model demonstrates a high level of accuracy in predicting crime rates over an extended period of time.
Nevertheless, the random forest model does include significant limitations. Due to the repetitive nature
of learning the model and creating the tree, the effectiveness of these processes is constrained by time
limitations. Furthermore, the model's performance is limited due to insufficient data and a scarcity of
accessible resources. The use of deep learning techniques to the issue of crime prediction has exhibited
promising preliminary outcomes.

9526
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

There are several advantages and merits associated with the utilization of advanced neural network
approaches, such as deep neural networks, partially generative neural networks, hierarchical recurrent
networks, and enhanced associative neural networks. As mentioned in prior studies (Huang et al., 2018;
Lin et al., 2018; Seo et al., 2018), the models under consideration exhibit higher AUROCs, effectively
capturing the temporal significance of crime incidents. Additionally, these models demonstrate the
ability to predict crime occurrences across various categories within different sectors of urban areas.
Furthermore, they facilitate the replication, transmission, and ongoing enhancement of the knowledge
model, while also offering a more objective foundation for comparative analysis.
CONCLUSION AND RECOMMENDATIONS
The convergence of electronic crimes and artificial intelligence heralds a new era in criminal law. This
paper has explored the dynamic landscape of criminal behavior in the digital age, examining the
evolution of electronic crimes, the integration of AI in criminal justice, legal responses to emerging
challenges, the delicate balance between individual rights and security, and relevant case studies. As
society grapples with these transformative forces, the role of criminal law must adapt to ensure justice,
security, and respect for human rights in an increasingly interconnected world.
As AI has a variety of tools that contribute to cyber security, there are still some areas that need
additional research that is more focused. There are certain inaccuracies associated with the models that
are currently available. As was just mentioned, there is a possibility of producing false alarms when
testing such models. In areas where research is still being done, there is also the possibility of malware
attacks that are carried out by themselves. Due to the design of increasingly complex models, there is
an ongoing requirement to provide training and support in order to enhance these models. However,
the same technologies that are used to prevent and detect cyber threats can also be used to activate
similarly dangerous attacks. Attacks on the security of computer systems or computer networks can
be made more flexible and critical when AI techniques are used. As a result, more robust models are
required to detect attacks of this critical nature.
The most applied algorithm utilized in crime prediction is random forest and naïve Bayes. Twenty-five
papers applied random forest and 20 papers applied naïve Bayes, whereas 17 papers utilized the
decision tree algorithm. In addition, most of the scientific papers used hybrid models that applied more
than one ML algorithm. Furthermore, the most utilized approach in the field of crime prediction is the
supervised learning approach, with a percentage of 31%. Moreover, 22% of the research papers
collected used a supervised and unsupervised approach. 10% applied unsupervised learning. The most
applied algorithm utilized in crime prediction is random forest and naïve Bayes. Twenty-five papers
applied random forest and 20 papers applied naïve Bayes, whereas 17 papers utilized the decision tree
algorithm. In addition, most of the scientific papers used hybrid models that applied more than one ML
algorithm. Furthermore, the most utilized approach in the field of crime prediction is the supervised
learning approach, with a percentage of 31%. Moreover, 22% of the research papers collected used a
supervised and unsupervised approach. 10% applied unsupervised learning. The most applied
9527
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

algorithm utilized in crime prediction is random forest and naïve Bayes. Twenty-five papers applied
random forest and 20 papers applied naïve Bayes, whereas 17 papers utilized the decision tree
algorithm. In addition, most of the scientific papers used hybrid models that applied more than one ML
algorithm. Furthermore, the most utilized approach in the field of crime prediction is the supervised
learning approach, with a percentage of 31%. Moreover, 22% of the research papers collected used a
supervised and unsupervised approach. 10% applied unsupervised learning.
The current increase in technology has led to an increase in the number of cyber attacks and security
threats. It is essential to have models that are not only flexible but also more robust and can be scaled
according to the data that is at hand. Throughout the course of this report, a variety of AI strategies for
detecting and preventing potential breaches of cyber security have been discussed. In addition to the
applications of artificial intelligence, there are also applications of deep learning that protect the
network's security. Combinations of techniques drawn from biology and those based on machine
learning can provide enhanced defense against these kindsof dangers. There are still new areas that
need to be developed so that knowledge of artificial intelligence can be used to improve the capabilities
of systems and create cyber security. One of these areas is the field of artificial intelligence.
REFERENCES
Al-Yaseen W. L., et el (2017) Multi Level Hybrid support vector machine and extreme learningmachine
based on modified K-means for intrusion detection. Future Generation Computational
Systems Vol 79.
Garg S., Batra S. (2018) Fuzzified Cuckoo based clustering technique for network anomalydetection,
Computational Electronic Engineering, Vol 71.
Hashemi H., Azmoodeh A., Hamzeh A. (2017) Graph embedding as a new approach for unknown
malware detection. Journal of Computational Viral Hacking Techniques Vol 13p.153-166.
Parisi A. (2019) Hands on Artificial intelligence for Cyber Security, Packt Publishing Ltd.
Sahingoz O. K., et el (2019) Machine Learning based phishing detection from URLs, ExpertSystem
Application, p.345-357.
Kim, S., Joshi, P., Kalsi, P. S., & Taheri, P. (2019a). Crime Analysis Through Machine Learning. 2018
IEEE 9th Annual Information Technology, Electronics and Mobile Communication
Conference, IEMCON 2018, 415–420. https://doi.org/10.1109/IEMCON.2018.8614828
Kim, S., Joshi, P., Kalsi, P. S., & Taheri, P. (2019b). Crime Analysis Through Machine Learning. 2018
IEEE 9th Annual Information Technology, Electronics and Mobile Communication
Conference, IEMCON 2018, 415–420. https://doi.org/10.1109/IEMCON.2018.8614828
Srivastava, A., Kundu, A., Sural, S., & Majumdar, A. K. (2008). Credit card fraud detection using Hidden
Markov Model. IEEE Transactions on Dependable and Secure Computing, 5(1), 37–48.
https://doi.org/10.1109/TDSC.2007.70228
Babakura, A., Sulaiman, M. N., & Yusuf, M. A. (2015). Improved method of classification algorithms for
crime prediction. Proceedings - 2014 International Symposium on Biometrics and
SecurityTechnologies, ISBAST 2014, 250–255.
https://doi.org/10.1109/ISBAST.2014.7013130
Nasridinov, A., Byun, J. Y., Um, N., & Shin, H. S. (2016). Application of data mining for crime analysis.
Lecture Notes in Electrical Engineering, 354, 503–508. https://doi.org/10.1007/978-3-662-
47895- 0_61
Nasridinov, A., Ihm, S. Y., & Park, Y. H. (2013). A decision tree-based classification model for crime
prediction. Lecture Notes in Electrical Engineering, 253 LNEE, 531–538.
https://doi.org/10.1007/978-94-007-6996-0_56
Saravanan, P., Selvaprabu, J., Arun Raj, L., Abdul Azeez Khan, A., & Javubar Sathick, K. (2021). Survey
on Crime Analysis and Prediction Using Data Mining and Machine Learning Techniques.
Lecture Notes in Electrical Engineering, 688, 435–448. https://doi.org/10.1007/978-981-
15-7241-8_31

9528
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Butt, U. M., Letchmunan, S., Hassan, F. H., Ali, M., Baqir, A., & Sherazi, H. H. R. (2020). SpatioTemporal
Crime HotSpot Detection and Prediction: A Systematic Literature Review. In IEEE Access (Vol.
8, pp. 166553–166574). Institute of Electrical and Electronics Engineers Inc.
https://doi.org/10.1109/ACCESS.2020.3022808
Kawthalkar, I., Jadhav, S., Jain, D., & Nimkar, A. V. (2020, February 1). A Survey of Predictive Crime
Mapping Techniques for Smart Cities. 2020 National Conference on Emerging Trends on
Sustainable Technology and Engineering Applications, NCETSTEA 2020.
https://doi.org/10.1109/NCETSTEA48365.2020.9119948
Albo, F. (n.d.). A Survey of Research into Artificial Neural Networks for Crime Prediction. Retrieved
May 21, 2021, from https://www.researchgate.net/publication/336116157
Zhao, X., & Tang, J. (2017). Modeling temporal-spatial correlations for crime prediction. International
Conference on Information and Knowledge Management, Proceedings, Part F1318, 497–506.
https://doi.org/10.1145/3132847.3133024
Zhao, X., & Tang, J. (2018). Crime in urban areas: A data mining perspective. In arXiv. arXiv.
https://doi.org/10.1145/3229329.3229331
Zhu, Q., Zhang, F., Liu, S., & Li, Y. (2019). An anticrime information support system design: Application
of K-means-VMD-BiGRU in the city of Chicago. Information and Management, 103247.
https://doi.org/10.1016/j.im.2019.103247
Shamsuddin, N. H. M., Ali, N. A., & Alwee, R. (2017). An overview on crime prediction methods. 6th
ICT International Student Project Conference: Elevating Community Through ICT, ICT-ISPC
2017, 2017-Janua, 1–5. https://doi.org/10.1109/ICT-ISPC.2017.8075335
Fredrick David, H. B., & Suruliandi, A. (2017). Survey on Crime Analysis and Prediction Using Data
Mining Techniques. ICTACT Journal on Soft Computing, 7(3), 1459–1466.
https://doi.org/10.21917/ijsc.2017.0202
Fredrick David KR, B. H., Benjamin Fredrick David, H., & Suruliandi, A. (2017). SURVEY ON CRIME
ANALYSIS AND PREDICTION USING DATA MINING TECHNIQUES Crime analysis and
prediction View project Content Based Image Retrieval View project SURVEY ON CRIME
ANALYSIS AND PREDICTION USING DATA MINING TECHNIQUES. Article in ICTACT Journal
on Soft Computing, 3. https://doi.org/10.21917/ijsc.2017.0202
Dilek S., Huseyin C., Aydm M. (Jan 2015) Applications of Artificial Intelligence techniques to combating
cybercrimes. International Journal of Artificial intelligence and applications,Vol 6 No 1.
McLaughlin N., et el (Mar 2017) Deep android malware detection, Conference on Data andApplication
Security and Privacy, p 301-308
Xu Z., Ray S., Subramanyan P., Malik S. (Mar 2017) Malware Detection using machine learningbased
analysis of virtual memory access patterns, Conference on Design, Automation & Test.
Patil P. (May 2019) Artificial Intelligence in Cyber Security, International Journal of research in
computer applications and robotics, Vol 4 Issue 5 p. 1-5.
Truong T. C., Diep B. Q., Zelinka I. (Mar 2020) Artificial Intelligence in Cyber Domain: Offense and
Defense, Symmetry 2020, Vol 12, Issue 410.

9529
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Table 1: Strength and Weakness of ML Models

ID Strength, weakness Model
A54 * Positive prediction is guaranteed by Negative binomial regression
count-based regression models.

A55 *Able to detect high potential hotspots at ensemble spatiotemporal pattern (ESTP)
which crime will have the highest
chances of being committed in the future
by setting a larger α
A56 *More robust to different time temporal-spatial correlations
granularities
A59 *Does not overfit the training data, Genetic programming – LSGP
resulting in near-optimal predicting on
previously encountered events.

A60 *More robust and has higher AUROCs Partially Generative Neural Networks (PGNN)
A62 *The feature space is heterogeneous and Random Forest
high-dimensional.

*It can handle with colinear features and

will not make any assumptions about
data on firsthand.
A63 *Captures the significance of crime Category Dependency Encoder + hierarchical
occurrences across time periods. Recurrent neural network + Multilayer
Perceptron (MLP)

* Predicts crime occurrences in many

categories in each sector of a city.

A66 * Allows for the replication, Deep Neural Networks

transmission, and continual
enhancement of the knowledge model,
as well as a more objective basis for
comparison.
A67 * Accurate geographical and functional feed-forward neural networks + Spatio-
affinity, activity preferences, and daily Temporal Embedding Similarity
schedules descriptions for places and
users
*Applications in a variety of situations
are possible.

A73 * Can be used to designate priority Risk Terrain Modeling

regions targeted by other nonpolice
agencies
A74 * Neural network-based methods might “Multi-View and Multi-Modal Spatial-
capture multidimensional spatial- Temporal learning (a multimodal pattern
temporal data's complex underlying fusion module + a hierarchical recurrent
structures in a nonlinear manner. framework)”
A75 * Effectively predict crime activity- “Fireflies-based fuzzy cognitive map neural
related features. networks (FFCM), Enhanced associative
neural networks (EANN)”

9530
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

A82 * Not only will this improve the accuracy * Bayesian hierarchical model
of mobility-based crime predictors, but it
also ensures that performance is
balanced across protected groups.

A83 * This algorithm works well with * K-NN

datasets that are nonlinear and have a
high order of complexity.
* Random forest
A84 (−) Learning the model is a time- * k-NN predictions
consuming iterative process, rendering
it unsuitable for use as an online tool.

(−) This method is ineffective for streets * demographic features

with little or no data.

*Weighted Multimodal Latent Dirichlet

Allocation (WMLDA)

A86 * Relatively resistant to overfitting, “logistic regression, support vector machines,

which is useful in this case because there decision trees, and random forests”
are only a few thousand grid cells to
learn from and almost a hundred POI
properties (SVM, Random Forest)
* Capable of using the fullest of available
attributes without overfitting - (SVM,
Random Forest)

A87 (−)Because of the nonlinear nature of * Extremely randomized Trees (Extra-Trees)

social data, LASSO prediction accuracy is
relatively low.
”Least Absolute Shrinkage and Selection
Operator (LASSO)”

A88 * No iteration steps are required for ”self-exciting point process models (SEPP)”
maximizing or decreasing likelihood or
cost functions, as in the EM method and
other machine learning techniques.

* The DDGF technique allows us to * data-driven Green's function

discover the buried causal effects of
criminal events.
* prospective hotspot maps
* Expectation Maximization
A93 * Accurate long-term crime forecasting Random Forests
in micro places with respect to other
common techniques
A106 * Simple to comprehend and implement J48 Decision Tree
in cybercrime investigations

9531
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

A110 * Less sensitive to spatiotemporal Random Forest Regression (RFR)

resolution, which allows it to perform
effectively in sparse areas.

Support Vector Regression (SVR)

Multi-Layer Perceptron Regression (MLPR)
A113 * Effectively avoid the problem of Random forest
overfitting
A118 * Secure, user-friendly and accurate with K-Means Clustering
predicting criminal activity

Utilized Datasets per Category

Country/State City Details Open From To % Freq.

Chicago * Na *Police Data Portal yes * 2001 * 2019 15% 19

Illinois

*Violent crimes * 2007 * 2012

* Police Data Portal * Jan 1st, * Dec 31st,

2017 2017

* Citizen Law *2013 *2017

Enforcement
Analysis and
Reporting (CLEAR)
India * Gujarat * Police department Yes Na Na 10% 13

* Bureau * National Crime

records

* Bengaluru * GVK Emergency

Management and
Research Institute
US * Cambridge - * Crime Analysis Na *Na *Na 9% 12
England Unit Police *Na *Na
Department *Jan 1st, *Dec 31st,
2006 2009
* Cheltenham * Crime data * Feb 10, * Aug 4,
- England 2010 2016

* Northeast * police department

- residential
burglary report

9532
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Country/State City Details Open From To % Freq.

* Southern *Department of
public safety at USC

*Baltimore *Baltimore Police

Department's
Victim Based Crime
Data
California * San *Homicide dataset Na *1981 *2014 9% 12
Francisco
* Mississippi
* Los Angeles *Na *Na *Na

*Police department *2014 *2016

Multi-media and Na * News sites, blogs, Na Na Na 9% 12

Social media RSS feeds
* Images of Knives,
guns, and blood
* Surveillance
videos
* Titter
* Open street map

New York * Police department Yes *2015 *2017 7% 9

* Crime data *July 1, *June 30,
* Crime data 2012 2013
*Jan 1, *Dec 31,
2014 2014
* 2014 *2015
Other Na * Point of Interest Na *Na *Na 6% 7

* Public Service *Na *Na

Complaint Data

* Taxi flow data *Na *Na

* News data *2011 *2019

* Global Terrorism *2015 *2016

Database (GTD)

9533
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Country/State City Details Open From To % Freq.

Gowalla Na Na Na *Feb 2009 * Oct 2010 5% 6

Foursquare

UCF Na *Crime dataset yes Na Na 5% 6

UCI * The Communities
and Crime
Unnormalized Data
Set

UK * London *Profiles (68 Na Na Na 4% 5

Borough metrics)
* police crime data

Brazil *Na * healthcare system Yes *Na Na 3% 4

*Fortaleza * July * Nov 2017
*Cearra 2016 * Nov 2017
* July
2016

Canada * Edmonton * Last 15 years Yes Na Na 3% 4

* Vancouver

Philippines *Misamis *Na Na *Na *Na 2% 2

Occidental *Manila Police *2012 *2016
*Manila District (MPD)

Bangladesh Na *Dhaka Na *June * May 2014 2% 2

Metropolitan Police 2013
(DMP)

Russia *Saint- * Crime records Private 1/1/2014 2/28/2017 1% 1

Petersburg

Hong Kong * Guangzhou * traffic violation Na 2012 2016 1% 1

Netherlands * Amsterdam * Police Department Na Na Na 1% 1

Pennsylvania * Na Na Na Na 1% 1
Philadelphia

Colorado * Denver Na Yes Na Na 1% 1

Portland * Oregon * Crime occurrences yes * March * December 1% 1

2012 2016

9534
Mustafa, A. H. The Future of Criminal Law in the Age of Electronic Crimes and Artificial Intelligence

Country/State City Details Open From To % Freq.

Australia * Queensland Na Na Na Na 1% 1

Taiwan * Taoyuan * Vehicle theft Na * Jan 2015 * April 2018 1% 1

Switzerland * Aargau *burglary incidents Na * Jan 14, * Jan 13, 1% 1

2014 2017

Mexico Na * Secretary-General Na * 2011 * na 1% 1

of National Public
Security
* The National
Council for the
Evaluation of Social
Development Policy

Belgium Na * local police Na * 2007 * 2017 1% 1

department

Argentina * Buenos Na Na *2016 *2019 1% 1

Aires

9535