دانلود رایگان مقاله انگلیسی + خرید ترجمه فارسی | |
عنوان فارسی مقاله: |
تشخیص موثر تقلب ها و کلاهبرداری های بانکی پیشرفته آنلاین در داده های به شدت نامتوازن |
عنوان انگلیسی مقاله: |
Effective detection of sophisticated online banking fraud on extremely imbalanced data |
|
مشخصات مقاله انگلیسی (PDF) | |
سال انتشار | 2013 |
تعداد صفحات مقاله انگلیسی | 27 صفحه با فرمت pdf |
رشته های مرتبط با این مقاله | مدیریت، مهندسی فناوری اطلاعات، مهندسی صنایع و مهندسی کامپیوتر |
گرایش های مرتبط با این مقاله | بانکداری، تجارت الکترونیک، اینترنت و شبکه های گسترده، هوش مصنوعی، داده کاوی |
چاپ شده در مجله (ژورنال) | وب جهان گستر – World Wide Web |
کلمات کلیدی | تشخیص تقلب، بانکداری آنلاین، الگوی مقابل، شبکه عصبی، داده کاوی |
ارائه شده از دانشگاه | دانشگاه تکنولوژی سیدنی، استرالیا |
رفرنس | دارد ✓ |
کد محصول | F995 |
نشریه | اسپرینگر – Springer |
مشخصات و وضعیت ترجمه فارسی این مقاله (Word) | |
وضعیت ترجمه | انجام شده و آماده دانلود |
تعداد صفحات ترجمه تایپ شده با فرمت ورد با قابلیت ویرایش | 32 صفحه با فونت 14 B Nazanin |
ترجمه عناوین تصاویر و جداول | ترجمه شده است ✓ |
ترجمه متون داخل تصاویر | ترجمه نشده است ☓ |
ترجمه متون داخل جداول | ترجمه نشده است ☓ |
درج تصاویر در فایل ترجمه | درج شده است ✓ |
درج جداول در فایل ترجمه | درج شده است ✓ |
درج فرمولها و محاسبات در فایل ترجمه | به صورت عکس درج شده است ✓ |
منابع داخل متن | به صورت عدد درج شده است ✓ |
کیفیت ترجمه | کیفیت ترجمه این مقاله متوسط میباشد |
فهرست مطالب |
چکیده
1. مقدمه
2. ویژگی های تقلب بانکی آنلاین و کارهای مرتبط
1.2. ویژگی های تقلب در بانکداری آنلاین
2.2. کار عمومی در تشخیص تقلب
3.2. تشخیص تقلب در بانکداری آنلاین
4.2. تشخیص کارت اعتباری تقلبی
5.2. تشخیص نفوذ به کامپیوتر
6.2. تشخیص تقلب از راه دور
3. بیان مسئله
4 . چارچوب سیستم
1.4. استخراج الگوی کنتراست
2.4. شبکه عصبی حساس به هزینه
3.4. جنگل تصمیم
5. رفتار بانکداری ماینینگ کانترست یا کانترست کاوی
1.5. چارچوب
5.2. مدل سازی رفتار پیچیده
3.5. گزیده ای از الگوهای رفتاری
6. خطر رفتار بانکی اینترنتی بر اساس مدل های ترکیبی
1.6. منطق
2.6. مدل منحصر به فرد به ثمر رساند خطر
1.2.6. امتیاز دهی در برابر الگوی ماینستر
2.2.6. امتیاز دهی توسط شبکه های عصبی حساس به هزینه
3.2.6. امتیاز دهی به جنگل تصمیم
3.6. ریسک با استفاده از مدل ترکیبی
7. آزمایش و ارزیابی
2.7. تنظیمات تجربی
3.7. ارزیابی عملکرد کلی
4.7. عملکرد رفتار مدل سازی کنتراست
8. نتیجه گیری
|
بخشی از ترجمه |
چکیده : 1. مقدمه |
بخشی از مقاله انگلیسی |
Abstract Sophisticated online banking fraud reflects the integrative abuse of resources in social, cyber and physical worlds. Its detection is a typical use case of the broad-based Wisdom Web of Things (W2T) methodology. However, there is very limited information available to distinguish dynamic fraud from genuine customer behavior in such an extremely sparse and imbalanced data environment, which makes the instant and effective detection become more and more important and challenging. In this paper, we propose an effective online banking fraud detection framework that synthesizes relevant resources and incorporates several advanced data mining techniques. By building a contrast vector for each transaction based on its customer’s historical behavior sequence, we profile the differentiating rate of each current transaction against the customer’s behavior preference. A novel algorithm, ContrastMiner, is introduced to efficiently mine contrast patterns and distinguish fraudulent from genuine behavior, followed by an effective pattern selection and risk scoring that combines predictions from different models. Results from experiments on large-scale real online banking data demonstrate that our system can achieve substantially higher accuracy and lower alert volume than the latest benchmarking fraud detection system incorporating domain knowledge and traditional fraud detection methods. 1 Introduction With the widespread use of increasingly advanced Internet technology [15, 47, 60], online banking (also called Internet banking) is emerging as a major channel for retail and business banking. In contrast, fraudulent online banking activities are becoming more and more sophisticated, seriously threatening the security and trust of online banking business. Online banking fraud has become a serious issue in financial crime management for all banks. It is becoming ever more challenging and leads to massive losses, due to the emergence and evolution of sophisticated online banking fraud, such as phishing scams, malware infection and ghost web sites. Effective and efficient detection of Internet banking fraud is regarded as a major challenge to all banks, and is an increasing cause for concern. An online banking fraud detection system can be a typical use case of the broadbased Wisdom Web of Things (W2T) [63–66] methodology. It has to timely gather multi-aspect data of online banking customers, including demographic data, online banking transaction data, credit card transaction data and other types of transaction data.These data will be transferred via the Internet/WWW and SEA-nets to an online banking customer data center. The data center provides a platform for the whole process of online banking fraud detection. It is a complete data cycle from acquisition of heterogeneous data, information, and knowledge in the physical world to the provision of active services in the cyber world to customers in the social world. Online banking customers (in the social world), things (in the physical world), and computer systems (in the cyber world) are integrated into an entity to realize their harmony and symbiosis by using an effective W2T data cycle. In this cycle, the process of fraud detection is one important task. Internet banking fraud exhibits certain sophisticated characteristics (see detailed discussions in Section 2.1): – suspicious customers are active and intelligent in conducting fraudulent banking activities, – fraudulent behavior is very dynamic, – fraud is hidden in diversified customer behavior, – fraud-related transactions are dispersed in highly imbalanced large data sets, and – the occurrences of fraud appear in a very limited time which requires real-time detection. The detection of online banking fraud needs to be instant, because it is very difficult to recover the loss if a fraud is undiscovered during the detection period. Most customers usually rarely check their online banking history regularly and are therefore not able to discover and report fraud transactions immediately after an occurrence of a fraud. This makes the possibility of loss recovery very low. In addition, all alerts generated from the detection system need to be manually investigated, which is very time-consuming. Online banking detection systems are therefore expected to have high accuracy, a high detection rate, and a low false positive rate for generating a small, manageable number of alerts in complex online banking business. The above characteristics and business requirements greatly challenge existing fraud detection techniques and data mining models for protecting credit card transactions, e-commerce, insurance, retail, telecommunication, computer intrusion, etc. These existing methods demonstrate poor performance in efficiency and/or accuracy when directly applied to online banking fraud detection [35]. For instance, credit card or telecommunication fraud detection often focuses on discovering particular behavior patterns of a specific customer or group, but fraud-related online banking transactions are very dynamic and appear very similar to genuine customer behavior. Some intrusion detection methods perform well in a dynamic computer environment, but they require a large amount of training data with complete attack logs as evidence. However, there is no obvious evidence to show whether an online banking transaction is fraudulent. A promising direction emerged recently that scrutinizes the difference between fraudulent and genuine behavior, and develops corresponding approaches for mining contrast patterns, for instance, contrast sets [6] and emerging patterns [24, 25, 52]. However, experiments of classic methods on real online banking data have shown that their accuracy is not very high because of the challenges in online banking fraud detection. In addition, according to the research in [61], contrast pattern mining is an NP hard problem, the time cost is expensive, especially when the number of attributes is large, and the threshold of minimal detection rate is small. Based on our experiments, the contrast pattern method in [24] does not perform efficiently in the online banking scenario. There are few papers about fraud control in online banking [35, 37, 44]. The mainstream online banking fraud detection systems rely on domain experts and knowledge to create rules for filtering suspicious transactions, which face critical problems, including very high false positive rates and low detection rates. More importantly, the adaptation of rules to fraud dynamics is fully dependent on domain expertise. This is very time-consuming, leaves the quality of fraud detection without sustainable control, and cannot support instant adjustment of rules. Most previous work treats events at different time points as independent and ignores the information incorporated in event sequences. In online banking, activity sequences are useful for differentiating fraudulent behavior from genuine behavior. An example is shown in Tables 1 and 2. Table 1 is a web page access sequence committed by a Trojan, while Table 2 is from a genuine transaction via a web browser. There are two contrasting features between these two sequences. One is that the fraud bypassed some web pages that are insignificant for submission of the transaction, such as homepage.aspx after login and the print page after the transfer confirmation. The other is that the transaction was completed within 3 seconds of login, which is too fast for a common online banking user to achieve via a web browser.Using the above data and business characteristics, this paper proposes an effective framework for detecting sophisticated Internet banking fraud efficiently. The main ideas, advantages and resulting contributions of this framework are as follows: – It is inspired by the theory of meta-synthetic engineering [11], M-Computing [12] and Wisdom Web of Things [66], and provides a systematic solution by synthesizing domain knowledge, experience learned in the rule-based detection system, advantages from multiple models, and refinement by domain experts. – It embeds systematic modules by selecting features based on information gain, extracting contrast behavior, building classifiers, generating an overall risk score for every online banking transaction, and identifying patterns of fraudulent behavior. This makes it a real time online banking fraud detection system that does not interfere with any existing online banking system or its service. – We not only construct sequence behavior information for identifying contrast patterns, but also propose a new method, a contrast vector, to integrate the sequential behavior contrast into the relational transaction database for mining more effective contrast patterns. – The system incorporates and integrates several data mining models, costsensitive neural network [67], contrast pattern mining, and decision forest. Because different models discover fraud and genuine behavior patterns from different angles, their combination [13] captures behavior patterns in a more comprehensive way. – Each model can be easily retrained over time to keep abreast of changes in fraud behavior. – Massive experiments in a major Australian bank show that our system and models have a higher detection rate and a lower false positive rate than any single classic data mining model, outperforming the existing rule-based system used in all major Australian banks. In addition, our system generates comparably good detection performance on highly imbalanced data sets and the modified contrast pattern mining model is efficient on real time data. The sequence behavior patterns discovered also provide more information about forensic evidence for fraud detection. The remainder of the paper is organized as follows. Section 2 describes the characteristics of online banking fraud in detail and presents an overview of related work on fraud detection. Section 3 gives a problem statement and definition of terminology, while Section 4 presents and explains the online banking fraud detection framework in detail. The method of contrast pattern mining with contrast vectors is introduced in Section 5 and the risk scoring method based on combined models is presented in Section 6. Experiment evaluation is discussed in Sections 7 and 8 provides conclusions and suggests future research directions. |