The research thesis is about fraud detection in knowledge base systems. We want to have a survey of fraud detection publications writing in ONTOLOGY, CASE BASED REASONING, EXPERT SYSTEM
In each domain we must have at least 4 publications. In the publications we discussed the importance, limitations of the publications at the end of the work we have a table of comparison for all of the publications we surveyed in all the three domains.
Survey of Fraud Detection Publications
Understanding fraud mechanisms to combat their consequences is, thus, a necessary challenge. Swindlers that use payment cards use several techniques to carry out their fraudulent activities. Not long afterward, banking research into morality in the banking sector was conducted, and proportionately, criminal practices became more complex. Fraud is described as accessing a facility, a product, or money by fraudulent methods that are becoming more popular across the world. Fraud as a felony can be difficult to identify at times.
Owners of credit cards carry out their preferred purchases by their preferences throughout a credit card’s use with a given account amount. Data needed for registering conducted transactions as a transaction in a graph type will be stored for each transaction performed by the customer (Ramaki et al., 2012). The system was introduced for identifying fraud in purchases involving a credit card. The data used for storage in an ontology graph like this is important to the detection process.
The system’s purpose was to establish a model for detecting credit card fraud centered on a semantic relationship between data stored for each consumer purchase, present it as an ontology graph, and store it in a patterns database. The recognition mechanism’s main benefit is the speed at which it can identify and save only worthy credit card consumers or other purchase behavior models in real-time (Fang et al, 2007). This method has restricted storage capacity for diverse data patterns due to the lack of comparable pattern maintenance, in addition to real-time fraud detection by running transactions.
Ontology-based fraud detection
Testing for unusual shifts in consumer behavior is one way to tackle cloning fraud. Using a set of data mining strategies explains the automated creation of consumer profiling tools to detect fraud. They use a rule-learning program to identify signals of fraud in a wide database of consumer purchases. The metrics are then used to construct a series of controls that profile typical consumer behavior and flag abnormalities. Finally, the monitors’ outputs are incorporated into a device that learns to integrate data to produce high-confidence warnings. Models for network intrusion detection are presented.
Fraud avoidance and early identification are now highly relevant priorities for the EU and its Member States. The legal core ontology, financial ontology, and fraud ontology were created as part of the FF POIROT project. We can use the technologies defined in the recommendation framework to construct the user behavior ontology for web applications, including e-commerce. Sufferers’ identity may be derived from users’ habits and aid suggestions in a recommendation framework. The machine uses personality knowledge to identify fraudulent behavior.
Electronic payment fraud prevention
Because of the increased usage of the Internet for shopping, electronic payment services have grown in popularity. This tool isn’t only for online transactions; more and more services are being built to render mobile money transfers simpler (Potamitis, 2013). As technology advances, the number of computers and processes available for conducting electronic transactions grows, whereas cash and checks decline. This is mostly because carrying credit cards or using mobile phones to make payments is much more convenient than carrying cash.
An individual who transfers money is interested in internet payments. Consumers may use a credit, debit, or prepaid card to withdraw money straight from their bank account, which is readily available online. The majority of people prefer the second alternative, especially when shopping online. With the rising e-commerce sector, this mode of payment is gaining prominence. The most popular means of electronic payment is the credit card. Credit, debit, and prepaid cards are the three forms of cards available (El Orche et al., 2018). They are usually made of plastic and have a magnetic line on the back. The customer passes the card to the vendor, who swipes it into a terminal or inserts the right information into a database. The invoice is then forwarded to the credit card company, which sends a notification to the broker that the transaction has been made.
One of the more important applications of ontologies is to distinguish domain information from organizational knowledge. The scheme specifies the role of electronic payment fraud rules based on its components’ electronic payment ontology specifications. It generates applications unaffected by-products and components.
Financial fraud has been around for a long time and may assume several different types. As a result of the increasingly evolving banking networks that have rendered it simpler for us to connect and perform financial transactions, the amount of fraud cases has risen, resulting in billions of dollars in damages per year across the world. To extract relevant information from such complicated datasets in the hunt for fraudulent operations, more than just a novel mathematical model is needed and the use of quick and effective Artificial Intelligence techniques (Sun & Finnie, 2004). Statistical methods such as univariate statistical simulations, Multiple Discriminant Analysis, and Linear Probability Models, Logistic Regression, and Probit analysis have been used to identify financial fraud. These approaches are accurate for limited sample sizes, and experience suggests that contingent and predictor variables have an inherent relationship.
Machine learning frameworks are rapidly being used in real-world applications, such as fraud detection. Complex structures, such as deep neural networks and ensembles, do the strongest. Despite their high accuracy, these models are not infallible, and forecasts can involve human domain experts to post-process them. However, as the model’s sophistication grows, human domain experts can find it more challenging to determine a forecast’s accuracy (Adedoyin, et al., 2016). Domain experts who post-process forecasts will learn from data on the model’s predictability in such situations.
Furthermore, this form of proof can help decide when a model is no longer trustworthy due to idea drift, which is especially valuable in fraud detection. The model’s own recorded confidence score is a simple measure of a prediction’s trustworthiness. On the other hand, Raw trust ratings are often under-calibrated, making them potentially deceptive for human domain experts. Since models aren’t flawless, and in the case of extremely imbalanced issues, such as fraud identification, even calibrated trust scores may be unreliable and thereby deceptive for fraud warning processing.
Case-based reasoning is a modern technology developed from a tiny and isolated research area to a burgeoning field of science. However, there has been no study into utilizing these methods to diagnose financial crime trends (Weerts, et al., 2019). The emphasis of the related literature on optimizing current approaches, the lack of sophistication of CBR research about the transaction implementation spectrum, and finally, the developed perception of the Financial Fraud Detection issue as one targeting precision optimization, rather than new forms of detecting and describing behavior trends, may all be contributing factors. This offers impetus to examine the efficacy of CBR methodology in identifying financial fraud.
Case-based logic (CBR) contributes significantly to the role of fraud identification. CBR programs may identify new cases by learning from sample examples of credit card use, and this method can respond to new fraud patterns as they arise. In the economic environment, CBR schemes have a range of benefits over other AI techniques. This includes the opportunity to improve from acquiring new cases over time without needing to introduce new laws or change current ones and decrease information elicitation commitment from dynamic and difficult transaction circumstances. the right to provide reasons when using previous events as precedent rather than creating new laws or changing old ones
Both active accounts’ authorization details will be modified, and the fraud model will be run for each active account. Accounts who fail the fraud exam will be written up in a report and forwarded to the bank’s fraud and protection section. Inspectors can then make communicate with the client and assess whether or not the latest conduct is legal. The account can be blocked in a matter of seconds. When handled improperly, blocking the account automatically decreases the financial institution’s liability exposure, but it has major customer care consequences (Behera & Panigrahi, 2017). The bank’s management agreed that calling the consumer as a first move would be the safest course of action. If the consumer agrees that theft has happened, the cardholder’s account will be blocked. One of the most critical predictions is that recent fraud practices would remain. The expense of having healthy accounts disrupted is not taken into consideration. This is largely due to the assumption that this is a tacit expense, and successful promotion of the fraud model’s efficacy for both parties will dramatically decrease this component of the total cost.
In all cases, the cardholder had no idea that his card’s replica owns thieves because the theft goes unnoticed. Consequently, the financial institution’s financial risk is considerably larger than it will be in the event of a missing or stolen passport, for example, where the cardholder is alerted nearly instantly. The issuance party is normally notified of a missing or compromised card during the first day, and a block is imposed on the account, restricting expenditure.
Since it offers an unambiguous specification of information and is flexible in a dynamic knowledge base, ontology is used to build an expert framework (Rajput et al., 2014). Domain information and certain laws to help logic make up the ontology-based expert method. Before ontology construction, data pre-processing is needed to choose unique data objects that can be translated into the ontology.
The development of a completely integrated system to combat money laundering is still a major challenge. Various heuristics-based AML recommendations are usable, and many commercial banks use them (Leonard, 1993). The expert framework focused on ontologies to identify fraudulent transactions. The usage of ontology improves the expert community’s efficiency since it needs less computation and allows the information base to be reused through implementations of related domains. The suggested ontology is made up of a transactions information base and SWRL laws. Generally, the publication’s importance is that it evaluates the key issue and threats of the domains along with the access of detection of frauds. The cost and the accuracy of the methods in the domains are the main limitations of the publications.
Evaluating case-based reasoning knowledge discovery in Fraud Detection
ONTOLOGY CASE BASED REASONING EXPERT SYSTEM
Domain experts may identify and apply fraud rules for a payment mechanism utilizing ontology in conjunction with different payment networks and data sources.
Different players and usage cases of electronic payment platforms were addressed, accompanied by a general summary of fraud management, and eventually, the recommended strategy for fraud reduction was introduced.
To apply the ontology and synchronize both, an interface between the ontology and the payment system data sources is being developed.
Another framework between the same ontologies has been included. Deception identification is classified into knowledge-based deception recognition, inference-based deception recognition, and hybrid deception recognition based on the rational interpretation of deception.
The proposed method would aid in the research and advancement of fraud and deceit detection in e-commerce and MAS.
The analytical study of deceit contributes to the classification of deception identification into three types: knowledge-based, inference-based, and hybrid deception recognition.
The proposed method would aid in the research and advancement of fraud and deceit detection in e-commerce and MAS. Because of the increased usage of information technologies and enterprises’ continuing proliferation, most financial transfers will now be conducted through electronic commerce systems.
Unfortunately, both legal consumers and fraudsters use these systems. Overall, the Expert Method model classifies at a lower degree of precision (91.78 percent).
However, the model accurately classifies 65.92 percent of the instances (381/578) as a fraud.
As a result, the cost of misclassification is poor, despite the lower accuracy standard.
Adedoyin, A., Kapetanakis, S., Petridis, M., & Panaousis, E. (2016). Evaluating Case-Based Reasoning Knowledge Discovery in Fraud Detection. In ICCBR Workshops (pp. 182-191).
Behera, T. K., & Panigrahi, S. (2017). Credit card fraud detection using a neuro-fuzzy expert system. In Computational intelligence in data mining (pp. 835-843). Springer, Singapore.
El Orche, A., Bahaj, M., & Alhayat, S. A. (2018, October). Ontology based on electronic payment fraud prevention. In 2018 IEEE 5th International Congress on Information Science and Technology (CiSt) (pp. 143-148). IEEE.
Fang, L., Cai, M., Fu, H., & Dong, J. (2007, May). Ontology-based fraud detection. In International Conference on Computational Science (pp. 1048-1055). Springer, Berlin, Heidelberg.
Leonard, K. J. (1993). Detecting credit card fraud using expert systems. Computers & industrial engineering, 25(1-4), 103-106.
Potamitis, G. (2013). Design and implementation of a fraud detection expert system using ontology-based techniques. University of Manchester (A dissertation submitted to the University of Manchester Giannis Potamitis School of Computer Science Table of Contents).
Ramaki, A. A., Asgari, R., & Atani, R. E. (2012). Credit card fraud detection based on ontology graph. International Journal of Security, Privacy and Trust Management (IJSPTM), 1(5), 1-12.
Rajput, Q., Khan, N. S., Larik, A., & Haider, S. (2014). Ontology based expert-system for suspicious transactions detection. Computer and Information Science, 7(1), 103.
Sun, Z., & Finnie, G. (2004, December). Experience based reasoning for recognising fraud and deception. In Fourth International Conference on Hybrid Intelligent Systems (HIS’04) (pp. 80-85). IEEE.
Weerts, H. J., van Ipenburg, W., & Pechenizkiy, M. (2019). Case-Based Reasoning for Assisting Domain Experts in Processing Fraud Alerts of Black-Box Machine Learning Models. arXiv preprint arXiv:1907.03334.
Such a cheap price for your free time and healthy sleep
All online transactions are done using all major Credit Cards or Electronic Check through PayPal. These are safe, secure, and efficient online payment methods.