Automated Fraudulent Phone Call Recognition through Deep Learning

Abstract

Several studies have shown that the phone number and call behavior generated by a phone call reveal the type of phone call. By analyzing the phone number rules and call behavior patterns, we can recognize the fraudulent phone call. The success of this recognition heavily depends on the particular set of features that are used to construct the classifier. Since these features are human-labor engineered, any change introduced to the telephone fraud can render these carefully constructed features ineffective. In this paper, we show that we can automate the feature engineering process and, thus, automatically recognize the fraudulent phone call by applying our proposed novel approach based on deep learning. We design and construct a new classifier based on Call Detail Records (CDR) for fraudulent phone call recognition and find that the performance achieved by our deep learning-based approach outperforms competing methods. Experimental results demonstrate the effectiveness of the proposed approach. Specifically, in our accuracy evaluation, the obtained accuracy exceeds 99%, and the most performant deep learning model is 4.7% more accurate than the state-of-the-art recognition model on average. Furthermore, we show that our deep learning approach is very stable in real-world environments, and the implicit features automatically learned by our approach are far more resilient to dynamic changes of a fraudulent phone number and its call behavior over time. We conclude that the ability to automatically construct the most relevant phone number features and call behavior features and perform accurate fraudulent phone call recognition makes our deep learning-based approach a precise, efficient, and robust technique for fraudulent phone call recognition.

Using Recurrent Neural Network Structure and Multi-Head Attention with Convolution for Fraudulent Phone Text Recognition

“…, n are used simultaneously. [1,2], [1,2,3], [1,2,3,4], [1,2,3,4,5], [1,2,3,4,5,6] five various multi-scale convolution kernels are selected for our experiments, and the number of convolution kernels is 150, 100, 75, 60, and 50 corresponding to them, respectively. The product of the convolution kernel of each scale and the number of kernels is equal to the output size of the hidden layer.…”

“…6, Tables 4 and 5, we can see that when the multi-scale convolution kernel of [1, 2, 3, 4, 5] is selected, Accuracy, Precision, Recall, and F1 have the highest values in the experiment. So [1,2,3,4,5] is selected as the type of kernel in the experiment. Through this convolution kernel, the model learns phrase knowledge of various lengths, and the multi-granularity local interaction improves the classification performance of the model.…”

“…Through this convolution kernel, the model learns phrase knowledge of various lengths, and the multi-granularity local interaction improves the classification performance of the model. When the multi-scale convolution kernel of type [1,2,3,4,5,6] is selected, the performance index decreases. This phenomenon is caused by the weak semantic connection between words that are too long in a phrase.…”

“…In the field of fraudulent phone recognition, traditional machine learning algorithms [2] are used by some researchers to identify fraudulent calls by extracting the difference between fraudulent calls and normal calls in terms of call frequency and call time. There are also researchers using deep learning methods [3] to replace artificial features with automatic features, which improves the recognition accuracy of fraudulent calls and ensures timeliness. Ying et al [4] proposed an efficient fraud phone detection framework based on parallel graph mining, which can automatically label fraudulent phone numbers to generate phone number trust values.…”

User identification and community exploration via mining big personal data in online platforms

“…In particular, malicious user detection of mobile phone numbers based on mobile phone data is one of the typical applications. Previous studies try to include as much information as possible for improving identification accuracy [99,78,141], without considerations of data privacy and efficiency. Third, for social network analysis on academic collaboration network, most previous works have not explore the differences among different scientific communities.…”