Abstract of Criminal Data Mining

The study presents the criminal data mining in the particular ways by which it can easily explores the types of the crimes that are generally occurring in our society. Due to the rapid growth of the crimes in the society it is necessary to overcome the ratio of the crimes. There are several ways strategies that are particularly applying by various authorizes in order to overcome such problem. Few of these are crucially technical and it needs the innovative firms of the information technology such as to detect the crime by using the data mining. The said study is conducted by using the Secondary source data collection in which the analysis of the case study is particularly utilized. Two case studies are discussing in this particular study. These are; Detecting Criminal Identity Deceptions: An Algorithmic Approach and Authorship Analysis in Cybercrime. Both of these case studies have been analyzed for exploring the criminal data mining in effective manners. All these tools are explaining in effective manners. The entire type of the crimes is discussing and explains in good ways by exploring the various kinds of the tools and techniques of the data mining.

Key words: Crimes, Data mining, Fraud, Theft and Arson

1.     Introduction of Criminal Data Mining

It can be seen from the reports that public security departments are extremely important for the country. This is because they are involved in maintaining economic prosperity and social stability. Moreover, their business is extensive that contains maintaining anti-criminal, social security, traffic and other different tasks of the society (Deshmukh, 2015). Furthermore, as technology is increasing there is the development of information construction departments. In such departments, complete public data can be accumulated easily in the public security information Centre for the analysis. But due to this technology, there is still a huge gap in the traditional functions for using this data. Moreover, there is some problem in the depth analysis of the prevention and prediction. (Li, 2016).

In nowadays, security is extremely important for the whole world. Moreover, it is one of the top priorities of the National Crime Records Bureau India. It can be noted that due to such security they are still facing extreme terrorist and insurgent activities in their country. This country is facing some terrorist incidents like bomb attacks. The first terrorist attack was on High court on 7th September 2011 that caused death of 12 people. Moreover, bomb attacks on 13th July 2011 in Mumbai in the crowded area. These bombs were placed in three different areas and due to this many people were killed. This was done on 13th February 2012 in New Delhi on Israeli diplomatic vehicle. The facts are showing that their crime graph is increasing continuously. This should be stopped because it will damage the overall reputation of the country in front of the world. Furthermore, if it is not possible to stop then it must be detected properly. For that case, a tool will be introduced that will use a reliable algorithm for detecting future crime in that area by data mining. (Sharma, 2014).

The next fact is that crime incidents are reported every day and they are increasing daily. This shows that the police department is the major organization for stopping such crime problems. Moreover, it is extremely difficult to find such a country with no crime. The current society is filled with a huge number of crimes. Only the police department can control the crime and increase security. Due to this people will feel safe and also maintain peace in society. If any government wanted to fight against the crime, then there is a requirement of law enforcement agencies. Furthermore, law enforcement agencies are required to collect data and evaluate it with perfection. Intelligence-led Policing is one of the most important models for proactive policing activity in a country. This department is involved in collecting basic data for informed decision making. It is completely based on the criminal analysis of the current situation. (Uzlov D. V., 2018).

For the automation of the criminal analysis process, there is a requirement of deductive interference procedures that will ensure all consequences and different procedures for extracting the data in the criminal environment. It will also evaluate the complex calculations in a perfect way. (MohammadReza Keyvanpoura, 2011). The relation between non-obvious criminal environment shows how the things are connected to each other. In the criminal analysis, the environment is often fizzy and show random number of connections. It is important to identify the relations between structured conditions. Figure 1 below demonstrates the multi valued logic under the descriptive components. The only thing is to identify the process and dynamic conditions. In the theory the operational search activity provides logic predictions and how the criminal activity proceeds.

Figure 1: Criminal environment Non-Obvious Relationships (Uzlov, Vlasov, & Strukov, 2018)

The next thing is that the communication trend of the modern era that includes instant messaging, chat server systems have become the most convenient method for sharing useful information. Due to this case, the criminals are misusing these opportunities for their illicit activities. The next fact is that from the National center for missing and exploited children, there are about one out of 7 children are facing online sexual solicitation in the United States of America. The next thing is that drug dealers are using chat rooms, these terrorists are using social media applications for their ideology (Awan, 2017).

It can be noted that these chat and IM system containing an archive feature that will save these conversations for later references. This shows that if the investigator is able to access this archived conversation in the chat servers or from computers then it will be extremely helpful in performing crime investigation. The next thing is that online communication content will reveal the lifestyle of the participants and their main social activities. Despite this, it is not easy to analyses the criminal activity from this chat data because it is a tedious and time-consuming task. Due to this case, the investigators are using different traditional search methods like tools or search engines to evaluate this data. (H. A. Shabat and N. Omar, 2015). The main aim of the data mining framework is to collect interpretable and instinctive data. This section deals with the proposed framework of this data mining. (Iqbal, 2019).

Figure 2: Highlight of the proposed framework. (Iqbal, 2019).

Figure 2 illustrates proposed framework for the confiscated machine and how the common concepts can relate to each other. The confiscated machines retrieve chat logs with the prepressed chat session. The framework consists of clique detector and concept miner.  In the current times, the concern about national security has been increasing just after the terrorist attacks on 11th September 2001. In the United States of America, there are about two intelligence agencies the FBI and CIA are involved in collecting and analyzing the criminal information about the terrorist and also their main activities. Due to this fact, the local law enforcement agencies are more alert in handling these activities. There is one of the huge challenges to law enforcement agencies and it is related to analyzing the large volume of data in an efficient way. But the data mining process is making it so much simple for these agencies because it can easily explore huge databases for users and organizations. There is a proper discussion about the various data mining techniques for analyzing the criminal data of the country in the context of law enforcement and intelligence analysis. It will also present four different case studies to illustrate the projects (Deshmukh, 2015).

Small intro for crime data mining

Data mining can be reviewed in two main dimensions. The first one is the types of crime and its security concerns. The next is techniques and approaches related to crime data mining.

2.1 Types of crime and security concerns of Criminal Data Mining

According to the fact, crime can be defined as “an act or the commission of an act that is forbidden, or the omission of a duty that is commanded by public law and that makes the offender liable to punishment by that law” (Webster Dictionary). In crime, there are different range of activities that may be range from simple violation of the traffic laws like illegal parking to high-quality crime like drug dealing or killing. In the given table1 below there is complete information about different types of crimes according to the public influence. Another thing is that local and national law enforcement authorities are facing the same kind of challenges. According to that case, data mining can be defined as the identification of the main structure in data and this structure is making patterns, predictive, statistical models of the data and showing the main relationship between them. (Kumar, 2015).

Table 1: Different types of crimes with relation to level of local law and level of national security[Ðl1]  (Osborne & Wernicke, 2013)

Different types of crimes	Level of local law enforcement	Level of national security
Violation of traffic	Personal injury, driving under influence, property damage, road rage, traffic accident
Crime related to sex	Sexual assault, sexual offenses, child molesting	Planned prostitution
Fraud	Counterfeiting and forgery fraud, embezzlement identity deception	Identity fraud, financial and money laundering fraud
Theft	Larceny, burglary, robbery, stolen property, motor theft,	Information hackers, national secrets theft and weapon information
Gang offenses	Sale and possession of drug	Trafficking of transnational drug
Arson	Building and apartments Arson
Cybercrime issues	Illegal trading, network instruction, hate crime, internet frauds, virus spreading, cyber-terrors, cyber-pornography, confidential information theft	Illegal trading, network instruction, hate crime, internet frauds, virus spreading, cyber-terrors, cyber-pornography, confidential information theft
Violent crimes	Heavily armed robbery, different assaults, criminal homicide	Terrorism

[Ðl2]  (Viano, 2016)

Research problem of Criminal Data Mining

With the passage of time the ratio of the crime is increasing gradually. Due to the rapid growth the security systems and needs to overcome such problem are also increasing (K. K. Sindhu1, 2012, ). The government and system must be adopting the particular ways by which the ratio of the crime can be stopped or decreased. There are the enormous ways that are adopted by the government of the area in order to stop and reduce the ratio of the crime. This particular study explores the various existing ways along with its tools and techniques in order to control the ratio of the crime. The said study is conducted to explain the data mining techniques for reducing the ratio of the crime on the society.

Research Objectives/ Motivation of Criminal Data Mining

There are the several research objectives and motivations of the study that must be attained in this particular study. All of these research objectives are required to highlight the major and most important points of the research study by which this study can be conducted. Few of these objectives and motivations are;

·         To examine the data mining technique’s potential that is required for the detection of the crime.

·         To explain the tools of the data mining by which the various decisional techniques can be supported for the experiments of the records of the crime.

·         To utilized the increased features for the selection methods as well as the importance of attributed factors for boosting the new tool performance.

·         To analyzed as well as interpret the tool performance for the various techniques.

The said study is contains on the major five sections. The first section is Introduction. This section introduce the topic in which the justification to select this topic for this research is explored. This part of the research study also explains the objectives and motivations of the study. The second section is the Literature Review. This section of the research explains about the related works of the various authors on this particular topic. The third section is Methods. The materials and methods that are particularly used to conducting this research study is explain under this section. The fourth section is Discussion. This is the one of the most important section of this research study because it explores the various techniques and tools for the identification of the crime. And the last section is Conclusion. This study concludes the major themes of the research study along with its major analysis.[Ðl3] 

Related work of Criminal Data Mining

The study on the previous related works is very important for both learning from the experiences of other people as well as addition some important things to existing information.

It has reviewed this literature in three different areas which are: data extraction, data mining as well as data focus. It also focused on the recent developments in applications of crime control on adopting the techniques of data mining to provide assistance to the process of crime investigation. One of the earlier projects is COPLINK, and the police department, as well as Arizona University, collaborate to extract the entities from the narrative records of police. As presented by Bruin, Cocx, Kosters, & Laros (2006),  [Ðl4] a tool to bring change in criminal behavior. It has also used the extracted factors involving the duration, frequency, nature as well as the seriousness for the comparison on the similarities among the criminal's pairs through the cluster data as well as the new distance measure consequently (Chen, Chung, Xu, & Chau, 2004).

A general framework was also presented by other researchers for the data mining of the crime that uses many methods and techniques of data mining such as knowledge discovery, association rule mining as well as link analysis to establish the new framework for crime data mining. The crime data mining framework will be able to determine different types of crimes. Furthermore, the efficiency and accuracy of the link analysis system, was also improved by Chau et al. through joining the different types of techniques including the heuristic approach, the shortest path algorithm as well as co-occurrence analysis to effectively determine the connections in the criminal network (Chauhan, 2017; Goyal, Bhatnagar, & Jain, 2020).

In the explanation of predictive policing, the predictive techniques as well as the analytical technique used to determine the criminals as well as they have been identified in a very effective manner by performing the same thing. We will have to manage the larger amount of data related to the crimes which are stored into the databases in warehouses due to the increase in the number of crimes in recent years. Furthermore, it is very difficult to identify as well as analyze the criminals and their data manually because the criminals have the knowledge about the technology as well as they are now up to date technologically. Due to this reason, the system of the police department should be updated as well as they also need to use advanced technologies to make police very effective to keep track of them. [Ðl5]  (Al-Janabi & Fatlawi, 2010)

The research study in this paper is organized in the following sequence and things. The writers of this research provide a brief review of the related studies and work on the crime analysis though some special forms of the algorithm as well as techniques. The five years of crime is analyzed in the Review Analysis section. Although, the future work, as well as the conclusion sections, are providing brief information or a summary of the paper is a very short way. Furthermore, future work is describing the new future work related to this study (Chauhan, 2017).

The other study which was conducted by Hussain, Durairaj, & Farzana (2012) [Ðl6] is also used the techniques of data mining to analyze the behaviors of criminals (Yadav, 2017; Kostyuchenko, Yuschenko, & Artemenko, 2020). The criminal investigation analysis tool is also proposed in this paper. To help to resolve the violent crime, the Central intelligence agency tool was also used in the community of law enforcement. From the witnesses, victims as well as the crime scenes, it was based on the evidence review. The comparison of the criminal records was performed from both behavioral perspectives as well as the investigative perspective. The insight was also provided in the investigative suggestions, unknown offenders as well as the strategies for the trial and interviews (K. K. Sindhu1, 2012, ).

Furthermore, solving crimes has also been the prerogative of the law enforcement specialist as well as criminal justice. Helping the detectives, as well as the law enforcement officers, have also been started by computer data analysis with the enhancing use of the computerized systems for tracking the major crimes to speed up the crime solution process. It will be looking by the researchers on how they can covert the information related to crime within the problem of data mining because the detectives can helpful in the speedy way to solve the crimes. It has also observed that the terminology of crime, a cluster is the crime group in the region geographically or the crime hotspot. While, the cluster in the data mining terminology is the similar data points such as the possible pattern of the crime (Wang Chunyu, 2016).

Therefore, the actual cluster or the cluster subset will also have one to one communication for the patterns of crime. Furthermore, the algorithms of clustering of data mining are similar and have equal weight to the task of determining record groups which are very similar among themselves from the rest of the data. It was committed by one or the same suspected groups that such clusters in their results will also useful to identify the crime spree. To find the variable for the best clustering, it is the further challenge of this research which is given information. After that, the detectives will have the presentation of these clusters to mine data deeply through the expertise of such domains (Tayal, 2015).

Thus, data mining is a very effective and powerful tool to make able investigators of criminals that may deficiency of training extensively as efficiently as well as to explore a large amount of data. Thousands of instructions can be processed by one computer within seconds as well as it can also save time. furthermore, running software as well as installing software may often cost less than training and hiring new staff. Moreover, computers are less susceptible to errors rather than humans because the computer has the ability to work in a speedy way, an effective way as well as for a long time. some crimes majorly concern with the police at the state level, the city as well as country levels. The local law of enforcement units and international or national agencies investigate other types of crimes (Sharmin, 2018).

Methods and approach of Criminal Data Mining

The research study is Descriptive in its research type. Usually, this kind of research is known as the case study research, it also includes; studying the particular situation for determining that how the general theory can be raised; how the existing theory has born for various particular situations. It includes; anthropological studies etc. The said study is conducted by using the secondary source of data collection. The secondary source data collection leads towards the preexisting data that has been used for various other projects.  The secondary source of data collection includes as the; It offers the information; if the existing and required data on the project or topic which is not directly or current application for the required chosen evaluation questions. It also includes the information which has already been protected, collected and processed another entity or researchers. It will disclose easily that which kinds of questions still required that is needed to be addressed; for this process, data has been yet collected.  This study explores the various case studies in order to highlight the major facts and figures of the study. These case studies explains the techniques of the data mining in good ways by introducing its numerous tools of the data mining.

Discussion of Criminal Data Mining
Case Studies of Crime Data Mining

Four case studies related to the crime data mining are presented in this paper based on analysis as well as crime characteristics technique which is discussed above that is the component of the ongoing project of COPLINK.

Detecting Criminal Identity Deceptions: An Algorithmic Approach

Police officers with the deceiving identities are provided often by the criminals to mislead the investigations of police such as using addresses, or fabricated birth and aliases.

The officers are prevented by a large number of data from examining matches of inexact manually. The taxonomy of criminal deceptions identity has been built based on the case study on the misleading criminal identities recorded in Tucson Police Department (TPD) that consisted of a name, date of birth, address as well as ID number deceptions. It is also identified that several criminals usually made some amendments to their identity information. For instance, they may change the name or change spelling or they also can change the digit sequence in the social security number. An algorithm approach was also developed based on the taxonomy to identify the misleading criminal identities automatically (Saha, Naskar, Dasgupta, & Dey, 2020).

The approach is using four identities which are: name, date of birth, address as well as social security number as compared to every corresponding field to make a pair of criminal record determination. Calculating the Euclidean Distance of agreement computed the total value of disagreement among two records that measures total field attributes. It will notice the deception in pair record when the overall disagreement value exceeds threshold value pre-determined and it is required in training processes. By using the same set of the actual identity records of criminals, an experiment was also conducted form Tucson Police Department (TPD) (Saini & Srivastava, 2019). The result is also showing that the generated algorithm identifies the identity deceptions of criminals around 94%.

Authorship Analysis in Cybercrime

The cybercrime investigation is made by their anonymous nature as well as a large number of very difficult cyberspace activities. The sheer amount of message, as well as constant changing author IDs, limit the conventional ways to deal with the issues rely on the effort manually. The authorship analysis framework is also proposed to track the identities automatically of cybercriminals by sending the message as they post. Three features of messages are identified including structural, style markers as well as content-specific features are also extracted and used to develop models based on these features under this framework. The experimental study is also conducted on Chinese and English messages data sets to evaluate the effectiveness of the framework by the small number of authors. Three inductive learning algorithms are also tested which are: backpropagation neural networks, decision tree as well as support vector machine algorithm. Average prediction accuracies around 80-90% for email messages are also achieved. Furthermore, 90 to 97% for newsgroup messages as well as 70 to 85% for Chinese Bulletin Board messages are achieved. It observed the important performance improvement when it added features on top style markers. The other two classifiers were outperformed by classified support vector machine (SVM) on all occasions (Yoo, Park, & Raman, Micro-Level Incident Analysis using Spatial Association Rule Mining, 2019). The promising future of using this framework was also indicated by the experimental results to address problems related to identity tracing.

Data Mining Techniques for Analyze of Crime Data

The data mining techniques are used to identify the unknown pattern of the related data in the mined data by comparison. For this purpose, several types of data mining tools are very popular to manipulate the algorithms of data mining. For the implementation of data mining, two different types of approaches are also discussed. The first approach is to copy data items from the warehouse. And the second approach is to mine data within the warehouse (Deshmukh, 2015; Yoo, Crime data warehousing and crime pattern discovery, 2019). Some other types of data mining techniques exist and popular which are given below.

Classification is mapping the data object within one or many predefined categories. Such kind of algorithms normally classifiers output such as in the formation of rules or decision trees. It will detect the ideal application intrusion to collect important abnormal as well as normal data of audit for the program or user then apply the categorization of the algorithm to learn classifiers that will identify the data audit as belonging to the abnormal class or normal class (Isafiade, 2016). The methods are based on comparison and support vector regression outperformed algorithms. These algorithms are combined with two support vector machines (SVM) classifiers (Ficara, Cavallaro, Meo, & Fiumara, 2020).

Clustering and segmentation are utilized for segmentation of the database into clusters or subset cluster depending on attributes set. It is a strategy to gather information into classes with indistinguishable qualities where the closeness of the intra-class is augmented or limited. Affiliation recognizes affinities/relationships among the assortment of information as reflected in the inspected records. An outcome is designed portraying rules of relationship in the information.[Ðl7]  (Hong, Liou, Wang, & Vo , 2014)

Decision Tree is a model of the prediction that can be observed as the tree. Furthermore, every branch is the question of classification, as well as every tree leaf, is the data set partition along with the classification. The data on every branch point is divided without excluding or losing the data. It conserves the large number of non-churners as well as churners while moving upward or downward on the tree. [Ðl8]  (Song & Lu, 2015)

Neural Networks are also known as the biological systems which have the ability to recognize the patterns as well as to learn or predict. It can also be called as the artificial neural networks are the programs by implementing the detection of the sophisticated pattern as well as the algorithms of machine learning on the computer to construct the model for prediction for the databases historically. [Ðl9]  (Barrett, Morcos, & Macke, 2019)

Link analysis identifies the relationships among all database fields. The insight will be provided by finding out the correlations in data audits to select the right of features system to detect intrusion. [Ðl10]  (Polites & Watson, 2009)

Sequence analysis is designing and establishing sequential patterns. We can be helped by these algorithms to get information for a better understanding of what kind of sequence of audit events based on time are commonly encountered. The significant elements of the behavior profile are frequent event patterns. [Ðl11]  (Zhang, Lin, Fournier-Viger, & Li, 2017)

Our applications of different types of techniques are described in this document for crim data mining. It has used the entity extraction method to recognize the person, vehicle, drug, personal properties as well as address automatically from the police narrative report. The concept space in the clustering techniques have also been utilized to automatically different related objects automatically within the records of crime and the objects can be: vehicles, persons, organizations, or any other real object name.

It has applied the Deviation detection to detect any kind of fraud, detection of network intrusion as well as crime analyses that include the activities of tracing abnormally. It has also used the classification to recognize spamming of email as well as to determine the authors or senders of those emails.

It has also used the string comparator for the identification of misleading information into the records of criminals. Furthermore, it has also used the social network analysis for the analysis of association as well as the roles of criminals between entities in the network of criminals. [Ðl12]  (GUPTA, 2014)

Table 2: Various techniques implemented for crime analysis using data mining techniques[Ðl13] 

P.No	Focus	Methods/Tools	Advantages	Future Work
1	The main focus on the prediction of activities of criminals as well as decrease crimes	·         ID3 Algorithm ·         Z-Crime Tool	Highlighting the technology of data mining for the prediction of activities of criminals (Haque, Weathington, & Guha, 2019)	For different techniques for selection feature as well as classification For mobile automatic reply, the feature will be added to emails
2	·         For the identification of various related patterns of crime, prediction of link, statistical analysis and hidden links of crime data ·         The main focus on the detailed analysis of the network of collaboration, dissolution, and collaboration of an organized group of crime	Algorithm of hidden link detection	·         Visual & intuitive ·         Predictive Approach	The predictive approach will use future work in the analysis of crime to help to stop the crime before the occurrence (Iqbal, 2019)

 

The typical techniques of data mining such as the prediction as well as classification, analysis of association, the analysis outlier as well as the analysis of clustering determine the patterns in the structured data. The patterns are identified by 3 new techniques from bothering unstructured as well as structured data. Various automated techniques of data mining have been developed by the researchers for both national security applications as well as local law enforcement (Sengottuvelan, 2015).

Clustering techniques group items of data classes along with the same characteristics for the minimization or maximization interclass similarity such as to determine the suspects that conduct the crimes into the same kind of technique or distinction between the related groups to distinctive groups. The set of predefined classes are not had by such kind of techniques for assigning the items. To automatically associate other objects such as organizations, persons as well as vehicles in the records of crime, the concept space algorithm based on the statistics are used by some researchers (Ludwig & Cook, 2004; Lithopoulos & Ruddell, 2016). Figure 3 illustrates the volume of crime data and relation with the incidence and complex crimes. The data mining is a powerful tool that is used by many investigators. The lack of extensive training can be explored by large databases and efficiency. The framework has encompassed the major crime under different types such as new intelligence data mining schemes and traditional methods (Blog. syncsort. com, 2018; Javatpoint. com, 2020).  

Association rule mining determines normally taking place the data sets of items in a database as well as shows the particular patterns the same as rules. It has applied this technique to the detection network intrusion for association rules deriving from the interaction history of the users. Furthermore, such a technique can be applied by the investigators to profiles of network intruders to assist in the identification of effective future attacks on the network (Kumar, 2015). There are nine types of mining rules missing. The frequently occurring sequences of items are found by the sequential pattern mining over the transaction set that occurred at different times. The intrusion pattern among them can also be identified by this approach in the network detection intrusion.

Conclusion of Criminal Data Mining

It is concluded that an overview of data mining of crime as well as four COPLINK case studies are presented in this paper. Crime data mining has a future promising future form encouraging results for increasing efficiency as well as the effectiveness of analysis of intelligence and criminal. It can explore several future directions in the young field. For instance, it can also develop more visual, intelligence, as well as intuitive criminal investigation technique for the network visualization as well as the crime patterns. For the identification of the crime patterns through techniques of clustering, we focus on data mining use. As the machine learning task, the contribution was to express the detection of crime patterns as well as in that way of using the techniques of data mining to provide assistance to police detectives to solve the critical cases. It is also identified the important attributes such as developed the scheme as well as using experts based on the learning of the semi-supervised methods for giving weight to important attributes. The proposed modeling technique was able to determine the patterns of crime from a large number of crimes by establishing the job for the crime detective easier.

The models will be created by us for predicting the active spots of crime as the future extension of this study that will assist in the police deployment at most expected of the crime to any given time of the window. It is to allow the most effective usage of the resources of the police. For the intellectual crime investigation, some very significant abilities of techniques of data mining were also leveraged by using the framework for multi-purpose in this research. The systematic approach was explained by the framework for using the MLP as well as SQM neural networks to classify as well as a cluster the data of crime.