Telecommunications Fraud Detection Research Articles

Feature difference-aware graph neural network for telecommunication fraud detection

With the continuous escalation of telecommunication fraud modes, telecommunication fraud is becoming more and more concealed and disguised. Existing Graph Neural Networks (GNNs)-based fraud detection methods directly aggregate the neighbor features of target nodes as their own updated features, which preserves the commonality of neighbor features but ignores the differences with target nodes. This makes it difficult to effectively distinguish fraudulent users from normal users. To address this issue, a new model named Feature Difference-aware Graph Neural Network (FDAGNN) is proposed for detecting telecommunication fraud. FDAGNN first calculates the feature differences between target nodes and their neighbors, then adopts GAT method to aggregate these feature differences, and finally uses GRU approach to fuse the original features of target nodes and the aggregated feature differences as the updated features of target nodes. Extensive experiments on two real-world telecom datasets demonstrate that FDAGNN outperforms seven baseline methods in the majority of metrics, with a maximum improvement of about 5%.

TFD-IIS-CRMCB: Telecom Fraud Detection for Incomplete Information Systems Based on Correlated Relation and Maximal Consistent Block

Telecom fraud detection is of great significance in online social networks. Yet the massive, redundant, incomplete, and uncertain network information makes it a challenging task to handle. Hence, this paper mainly uses the correlation of attributes by entropy function to optimize the data quality and then solves the problem of telecommunication fraud detection with incomplete information. First, to filter out redundancy and noise, we propose an attribute reduction algorithm based on max-correlation and max-independence rate (MCIR) to improve data quality. Then, we design a rough-gain anomaly detection algorithm (MCIR-RGAD) using the idea of maximal consistent blocks to deal with missing incomplete data. Finally, the experimental results on authentic telecommunication fraud data and UCI data show that the MCIR-RGAD algorithm provides an effective solution for reducing the computation time, improving the data quality, and processing incomplete data.

CDR2IMG: A Bridge from Text to Image in Telecommunication Fraud Detection

CDR2IMG: A Bridge from Text to Image in Telecommunication Fraud Detection

Utilizing Prediction Intervals for Unsupervised Detection of Fraudulent Transactions: A Case Study

Money laundering operations have a high negative impact on the growth of a country’s national economy. As all financial sectors are increasingly being integrated, it is vital to implement effective technological measures to address these fraudulent operations. Machine learning methods are widely used to classify an incoming transaction as fraudulent or non-fraudulent by analyzing the behaviour of past transactions. Unsupervised machine learning methods do not require label information on past transactions, and a classification is made solely based on the distribution of the transaction. This research presents three unsupervised classification methods: ordinary least squares regression-based (OLS) fraud detection, random forest-based (RF) fraud detection and dropout neural network-based (DNN) fraud detection. For each method, the goal is to classify an incoming transaction amount as fraudulent or non-fraudulent. The novelty in the proposed approach is the application of prediction interval calculation for automatically validating incoming transactions. The three methods are applied to a real-world dataset of credit card transactions. The fraud labels available for the dataset are removed during the model training phase but are later used to evaluate the performance of the final predictions. The performance of the proposed methods is further compared with two other unsupervised state-of-the-art methods. Based on the experimental results, the OLS and RF methods show the best performance in predicting the correct label of a transaction, while the DNN method is the most robust method for detecting fraudulent transactions. This novel concept of calculating prediction intervals for validating an incoming transaction introduces a new direction for unsupervised fraud detection. Since fraud labels on past transactions are not required for training, the proposed methods can be applied in an online setting to different areas, such as detecting money laundering activities, telecommunication fraud and intrusion detection.

BTG: A Bridge to Graph machine learning in telecommunications fraud detection

Telecommunications fraud runs rampant recently around the world. Therefore, how to effectively detect fraudsters has become an increasingly challenging problem. However, previous studies either assume that the samples are independent of each other and use non-graph methods, or use local subgraphs with good connectivity for graph-based anomaly detection. Few prior works have performed graph-based fraud detection on real-world Call Detail Records (CDR) meta data sets with sparse connectivity. To solve this problem, we propose an end-to-end telecommunications fraud detection framework named Bridge To Graph (BTG). BTG leverages the subscriber synergy behavior to reconstruct connectivity, which bridges the gap between sparse connectivity data and graph machine learning. Concretely, we extract multi-model features from meta data and perform Box–Cox transformation first. Then, aiming at the sparse connectivity of real-world CDR meta data, the graph is reconstructed through dimensionally selectable link prediction of node similarity. Finally, the reconstructed graph and node features are input into the graph machine learning module for node embedding representation learning and fraud node classification. Comprehensive experiments on the real-world telecommunications network CDR data set show that our proposed method outperforms the classic methods in many metrics. Beyond telecom fraud detection, our method can also be extended to anomaly detection scenarios with no graph or sparse connectivity graph.

Voice analysis system for detection of vishing using deep learning

For many years, telecom fraud has caused significant financial harm to Indian telecommunications users. Traditional approaches for identifying telecom fraud frequently center on boycotting of faux phone numbers. Attackers, on the other hand, could merely escape discernment by modifying their phone numbers, which is fairly straightforward with VoIP (Voice over IP).To address this issue, this method detects telecommunication fraud supporting the substance of a spoken language rather than merely the caller's sign. This paper collects chronicles of telecommunication deceit, above all, from press sources and social platforms. To make datasets, our planned model utilizes machine learning techniques to look at knowledge and opt for high-quality descriptions from antecedently gathered knowledge. After that, Natural language processing is employed to draw out characteristics from the text- based data. Then, for extra telecommunication fraud detection, criteria to acknowledge identical material within the same call are formed. To spot telecommunication fraud online, the system provides an associate degree android application that will be loaded on a customer's smartphone. Once an associate degree incoming fraud decision is answered, the program will dynamically measure the call's contents to sight fraud. Our findings demonstrate that the system will effectively safeguard clients.

Recognition of Human Emotions by Voice in the Fight against Telephone Fraud

Advances in communication technologies have made communication between people more accessible. In the era of information technology, information exchange has become very simple and fast. However, personal and confidential information may be available on the Internet. For example, voice phishing is actively used by intruders. The harm from phishing is a serious problem all over the world, and its frequency is growing. Communication systems are vulnerable and can be easily hacked by attackers using social engineering attacks. These attacks are aimed at tricking people or businesses into performing actions that benefit attackers, or providing them with confidential data. This article explores the usefulness of applying various approaches to training to solve the problem of fraud detection in telecommunications. A person's voice contains various parameters that convey information such as emotions, gender, attitude, health and personality. Speaker recognition technologies have wide areas of application, in particular countering telephone fraud. Emotion recognition is becoming an increasingly relevant technology as well with the development of voice assistant systems. One of the goals of the study is to determine the user model that best identifies fraud cases. Machine learning provides effective technologies for fraud detection and is successfully used to detect such actions as phishing, cyberbullying, and telecommunications fraud.

Feature Selection Using Multivariate Adaptive Regression Splines in Telecommunication Fraud Detection

Feature selection determines the most significant features for a given task while rejecting the noisy, irrelevant and redundant features of the dataset that might mislead the classifier. Besides, the technique diminishes the dimensionality of the attribute of the dataset, thus reducing computation time and improving prediction performance. This paper aims to perform a feature selection for classification more accurately with an optimal features subset using Multivariate Adaptive Regression Splines (MARS) in Spline Model (SM) classifier. A comparative study of prediction performance was conducted with other classifiers including Decision Tree (DT), Neural Network (NN) and Support Vector Machine (SVM) with similar optimal feature subset produced by MARS. From the results, the MARS technique demonstrated the features reduction up to 87.76% and improved the classification accuracy. Based on the comparative analysis conducted, the Spline classifier shows better performance by achieving the highest accuracy (97.44%) compared to other classifiers.

Detecting telecommunication fraud by understanding the contents of a call

Telecommunication fraud has continuously been causing severe financial loss to telecommunication customers in China for several years. Traditional approaches to detect telecommunication frauds usually rely on constructing a blacklist of fraud telephone numbers. However, attackers can simply evade such detection by changing their numbers, which is very easy to achieve through VoIP (Voice over IP). To solve this problem, we detect telecommunication frauds from the contents of a call instead of simply through the caller’s telephone number. Particularly, we collect descriptions of telecommunication fraud from news reports and social media. We use machine learning algorithms to analyze data and to select the high-quality descriptions from the data collected previously to construct datasets. Then we leverage natural language processing to extract features from the textual data. After that, we build rules to identify similar contents within the same call for further telecommunication fraud detection. To achieve online detection of telecommunication frauds, we develop an Android application which can be installed on a customer’s smartphone. When an incoming fraud call is answered, the application can dynamically analyze the contents of the call in order to identify frauds. Our results show that we can protect customers effectively.

Perspective analysis of telecommunication fraud detection using data stream analytics and neural network classification based data mining

With the growing advancements in technology, the lives of the people have become easier and convenient, but at the same time it also mushrooms sophisticated practices through which the fraudsters can infiltrate an organization. Telecommunication industry, being one of the major sectors in the world, is also infiltrated by frauds. Telecommunication fraud is a combination of variety of illegal activities like unauthorized and illegitimate access, subscription identity theft and international revenue share fraud etc. Frauds have proven to be detrimental to the prosperity of a company and impacts customer relations and shareholders. This paper presents the implementation details for detecting telecommunication fraud using Data Stream Analytics and Neural Network classification based Data Mining. For detection using Data Stream Analytics, Event Hub and Stream Analytics components of Microsoft Azure have been used whereas for detection using Data Mining Neural Network Pattern Recognition tool as well as a self-coded algorithm in Matlab has been used. Based on the results, the accuracy of both the techniques have been compared and the situations for selection of a suitable technique based on the user requirements and the flow of data has been narrowed down. The findings can elucidate upon other cloud analytics systems and provide a basis for big data analytics and mining.

Employing Kullback-Leibler divergence and Latent Dirichlet Allocation for fraud detection in telecommunications

In this paper, a method for telecommunications fraud detection is proposed. The method is based on the user profiling using the Latent Dirichlet Allocation LDA. Fraudulent behavior is detected with use of a threshold-type classification algorithm, allocating the telecommunication accounts into one of two classes: fraudulent account and non-fraudulent account. The paper provides also a method for automatic threshold computation. The accounts are classified with use of the Kullback-Leibler divergence KL-divergence. Therefore, we also introduce three methods for approximating the KL-divergence between two LDAs. Finally, the results of experimental study on KL-divergence approximation and fraud detection in telecommunications are reported.

A probabilistic approach to fraud detection in telecommunications

In this paper, a method for telecommunications fraud detection is proposed. The method is based on the user profiling utilizing the Latent Dirichlet Allocation (LDA). Fraudulent behavior is detected with use of a threshold-type classification algorithm, allocating the telecommunication accounts into one of two classes: fraudulent account and non-fraudulent account. The paper provides also a method for automatic threshold computation. The accounts are classified with use of the Kullback–Leibler divergence (KL-divergence). Therefore, we also introduce three methods for approximating the KL-divergence between two LDAs. Finally, the results of experimental study on KL-divergence approximation and fraud detection in telecommunications are reported.

Fraud Detection in Telecommunications and Banking: Discussion of Becker, Volinsky, and Wilks (2010) and Sudjianto et al. (2010)

The papers by Becker, Volinsky, and Wilks (2010) and Sudjianto et al. (2010) provide overviews of fraud detection in telecommunications and banking respectively. Becker, Volinsky, and Wilks (2010) describe their experience in developing a system implemented at AT&T, and Sudjianto et al. (2010) review methods which can be used for fraud detection, especially in the retail banking sector. These two papers are put in a broader context, and compared with my own experience.

Fraud Detection in Telecommunications: History and Lessons Learned

Fraud detection is an increasingly important and difficult task in today’s technological environment. As consumers are putting more of their personal information online and transacting much more business over computers, the potential for losses from fraud is in the billions of dollars, not to mention the damage done by identity theft. This paper reviews the history of fraud detection at AT&T, one of the first companies to address fraud in a systematic way to protect its revenue stream. We discuss some of the major fraud schemes and the techniques used to address them, leading to generic conclusions about fraud detection. Specifically, we advocate the use of simple, understandable models, heavy use of visualization, and a flexible environment and emphasize the importance of data management and the need to keep humans in the loop.

Mini-Issue on Anomaly Detection

This issue of Technometrics includes three articles that highlight interesting applications of statistical methods for detecting unusual circumstances, or anomalies. The articles explore the methods that have been developed to address these problems and describe additional challenges that lie ahead. Our purpose in publishing these articles is to inform readers of some of the exciting applications and research in anomaly detection outside the bounds of traditional industrial statistics. Vijay Nair recruited and edited these articles and I want to thank Vijay for bringing this important topic to Technometrics. We begin with two papers on fraud detection. The first paper, by Sudjianto et al., provides an overview of the problems faced by the financial industry. The goal is to identify fraudulent activity within a massive stream of transaction data, which could be from a bank account, a credit card, an organization, etc. The paper describes some of the statistical techniques currently used, and the research challenges in the field. The second paper, by Becker, Volinsky, and Wilks, reviews fraud detection in telecommunications, with a focus on work that has been done at AT&T. It discusses statistical techniques, visualization tools, computing environment, and data management issues that have been important in improving fraud detection capability. These two papers are followed by a commentary from David Hand, who expands on the contributions and provides his own perspectives from involvement in a number of related applications. The third paper, by Shmueli and Burkom, deals with modern biosurveillance, where the goal is early detection of the outbreak of diseases. The paper describes various types of new data sources, the current state of monitoring, and the statistical challenges in modern biosurveillance. There are many other applications areas where anomaly detection issues arise. Some examples include early detection of warranty problems, intrusion detection in computer networks, and fault detection in sensor networks. While all of these can be viewed at a basic level as a problem of monitoring underlying processes and detecting changes, the issues and statistical challenges are quite different from traditional statistical process monitoring. We are pleased to use this special collection of articles to expose Technometrics readers to some of the fascinating problems and statistical research questions currently stimulated by anomaly detection.

An application of supervised and unsupervised learning approaches to telecommunications fraud detection

This paper investigates the usefulness of applying different learning approaches to a problem of telecommunications fraud detection. Five different user models are compared by means of both supervised and unsupervised learning techniques, namely the multilayer perceptron and the hierarchical agglomerative clustering. One aim of the study is to identify the user model that best identifies fraud cases. The second task is to explore different views of the same problem and see what can be learned form the application of each different technique. All data come from real defrauded user accounts in a telecommunications network. The models are compared in terms of their performances. Each technique’s outcome is evaluated with appropriate measures.

Employing Latent Dirichlet Allocation for fraud detection in telecommunications

Fraud detection in the telecommunications industry requires the identification of a small fraction of fraudulent calls from the high volume of call traffic. This presents a significant research challenge in the design of efficient and effective algorithms to combat telecommunications fraud. This paper employs Latent Dirichlet Allocation (LDA) to build user profile signatures and assumes that any significant unexplainable deviations from the normal activity of an individual user is strongly correlated with fraudulent activity. The user activity is represented as a probability distribution over call features which surmises the user’s calling behaviour. This probability distribution is derived from LDA which can accurately describe user profiles by combining different classes of distributions. To score calls we compare the likelihood of the user generating a call versus a fraudster generating the same call. Our experiments demonstrate that using such a probability distribution and employing even a rough profile of a fraudster’s activity ameliorates the detection of fraudulent calls. Our method is computationally efficient and can scale up to a realistic real time detection.

Computational Methods for Dynamic Graphs

This article considers problems that can be characterized by large dynamic graphs. Communication networks provide the prototypical example of such problems where nodes in the graph are network IDs and the edges represent communication between pairs of network IDs. In such graphs, nodes and edges appear and disappear through time so that methods that apply to static graphs are not sufficient. Our definition of a dynamic graph is procedural. We introduce a data structure and an updating scheme that captures, in an approximate sense, the graph and its evolution through time. The data structure arises from a bottom-up representation of the large graph as the union of small subgraphs centered on every node. These subgraphs are interesting in their own right and can be enhanced to form what we call communities of interest (COI). We discuss an application in the area of telecommunications fraud detection to help motivate the ideas.