Social networks play a dominant role in connecting communication devices, which play as essential role in exchanging a large amount of interpersonal data. As a result of exchange of messages, they leave behind some traces, which help in identifying the nature of the network. So, these are helpful in detecting crime users. The algorithms like SVM, Bayesian linear regression will not help in finding out the crime network. It also results in less accuracy for higher amounts of data. So, developing the trace learning system in GAN, which is a higher order of deep learning neural network, larger neural network dataset will be fed into the model, which has digital traces into the neural network done through proposed CrimedetGAN. A trained accuracy system model automatically identifies the digital traces which have been left in the crime natured social network. Experimenting with existing GAN frameworks, namely MaliGAN, seqGAN, LeakGAN, proposed CrimedetGAN came with a test score accuracy of 91.23% on the coherence NLP testing in tracing the relevant data fields for the given input datasets.
TopIntroduction
Crime occurrences reported to law enforcement include the event date and time, the geographical location, and a brief narrative description of the event. Recently, crime theme models (Kuang, Brantingham, & Bertozzi, 2017) for crime report narratives were established to help categories occurrences into more detailed and accurate categories. These models are more coherent than human-defined categories and may result in more accurate geographic risk estimations (Pandey & Mohler, 2018). Crime event text data also appears on social media, such as Twitter (Wang, Gerber, & Brown, 2012), and machine learning algorithms have been developed to recognize, categories, and integrate such events in spatiotemporal forecasts. Recently, neural network-based representations in the form of restricted Boltzmann machines (RBM) for crime report data have been developed (Zhu & Xie, 2019).
Crime detection in social networks using Generative Adversarial Networks (GANs) is an emerging research area that utilizes the power of GANs to identify and predict criminal activities in online platforms. GANs offer a unique approach by generating synthetic criminal data that closely resembles real criminal behaviors, enabling more accurate and comprehensive training of crime detection models.
One application of GANs in crime detection is the generation of synthetic criminal profiles or posts that mimic the characteristics of actual criminal activities. These synthetic data points can be used to train machine learning models to recognize patterns, anomalies, and indicators of criminal behavior. By leveraging GANs, crime detection systems can capture the diversity and complexity of criminal activities, including fraud, cyberbullying, hate speech, and illicit activities, among others. The use of GANs in crime detection has several advantages. GANs can capture the underlying structure and dynamics of social networks, allowing for the detection of subtle and evolving criminal behaviors that may go unnoticed by traditional methods. They can also adapt and learn from unlabeled data, enabling the identification of emerging criminal patterns. Moreover, GAN-based crime detection models can improve their performance over time as they receive feedback and learn from real-time data.
However, there are challenges associated with using GANs for crime detection in social networks. Acquiring labeled criminal data for training GANs can be challenging, as it requires access to real criminal activities and may raise privacy concerns. Additionally, the scalability of GAN-based models in handling the vast amount of data generated in social networks is a significant consideration. Model performance, interpretability, and ethical considerations also need to be addressed to ensure the reliable and responsible use of GANs in crime detection.
Overall, crime detection in social networks using GANs has the potential to enhance the accuracy and effectiveness of identifying and preventing criminal activities in online platforms. By leveraging the power of GANs, researchers and practitioners aim to develop advanced crime detection systems that can contribute to creating safer and more secure online environments.
GANs are composed of two networks: one network that converts random Gaussian vectors from a latent space to artificial observations (such as artificial pictures or text reports), and another network that learns to discriminate between real and artificial data. The GAN learns an accurate representation of the data after successful (adversarial) training, and artificial and genuine instances may become unrecognisable. In this study, we aim to assess various approaches for generative modelling of crime text data.
Figure 1. Model prediction mechanism of GAN