Discovering child sexual abuse material creators' behaviors and preferences on the dark web

BACKGROUND

Producing, distributing or discussing child sexual abuse materials (CSAM) is often committed through the dark web to stay hidden from search engines and to evade detection by law enforcement agencies. Additionally, on the dark web, the CSAM creators employ various techniques to avoid detection and conceal their activities. The large volume of CSAM on the dark web presents a global social problem and poses a significant challenge for helplines, hotlines and law enforcement agencies.

OBJECTIVE

Identifying CSAM discussions on the dark web and uncovering associated metadata insights into characteristics, behaviors and motivation of CSAM creators.

PARTICIPANTS AND SETTING

We have conducted an analysis of more than 353,000 posts generated by 35,400 distinct users and written in 118 different languages across eight dark web forums in 2022. Out of these, approximately 221,000 posts were written in English and contributed by around 29,500 unique users.

METHOD

We propose a CSAM detection intelligence system. The system uses a manually labeled dataset to train, evaluate and select an efficient CSAM classification model. Once we identify CSAM creators and victims through CSAM posts on the dark web, we proceed to analyse, visualize and uncover information concerning the behaviors of CSAM creators and victims.

RESULT

The CSAM classifier, based on Support Vector Machine model, exhibited good performance, achieving the highest precision of 92.3 % and accuracy of 87.6 %. While, the Naive Bayes combination is the best in term of recall, achieving 89 %. Across the eight forums in 2022, our Support Vector Machine model detected around 63,000 English CSAM posts and identified near 10,500 English CSAM creators. The analysis of metadata of CSAM posts revealed meaningful information about CSAM creators, their victims and social media platforms they used. This included: (1) The topics of interest and the preferred social media platforms for the 20 most active CSAM creators (For example, two top creators were interested in topics like video, webcam and general content in forums, and they frequently used platforms like Omegle and Skype); (2) Information about the ages and nationalities of the victims typically mentioned by CSAM creators, such as victims aged 12 and 13 with nationalities including British and Russian; (3) social media platforms preferred by CSAM creators for sharing or uploading CSAM, include Omegle, YouTube, Skype, Instagram and Telegram.

CONCLUSION

Our CSAM detection system exhibits high performance in precision, recall, and accuracy in real-time when classifying CSAM and non-CSAM posts. Additionally, it can extract and visualize valuable and unique insights about CSAM creators and victims by employing advanced statistical methods. These insights prove beneficial to our partners, i.e. national hotlines and child agencies.

Automated evaluation of Tuberculosis using Deep Neural Networks

INTRODUCTION: Tuberculosis (TB) is a chronic, progressive infection that often has a latent period after the initial infection period. Early awareness from those period to have better prevention steps becomes an indispensable part for patients who want to lengthen their lives. Hence, applying cutting-edge technologies to support the medical business domain plays a key role in improving speed and accuracy in methods of diagnosis. Deep Neural Network-based technique (DNN) is one of such methods which offers positive results by leveraging the advantages of analyzing deeply the data, especially image data format via tons of deep layers of the neural networks. Our study was wrapped up by objectively assessing the performance of modern Deep Neural Network approaches and suggesting a model offering good results in terms of the selected metrics as defined later. In order to achieve optimized results, the chosen model must adapt well to the datasets but require less hardware and computational resources.OBJECTIVES: Our objective is to pick up and train a Deep Neural Network architecture which is highly trusted and flexibly fitted and applied to various datasets with minimum configurations. This will be used to produce good predictions based on the input data which are Chest X-ray images retrieved from the published datasets.METHODS: We have been approaching this problem by using the recognized datasets which have already been published before, then resizing them to the consistent input data for training purposes. In terms of Deep Neural Networks, we picked up VGG16 as the baseline network architecture, then use other ones which are state-of-the-art networks for comparison purposes. After all, we recommend the neural network architecture offering the most positive results based on accuracy and recall measurements. So that, this network architecture will show flexibility when fitting into diverse datasets representing different areas in the world that suffered from Tuberculosis before.RESULTS: After conducting the experiments, we observed that the Mobilenet model produced great results based on the predefined metrics for most of the proposed datasets. It shows the versatility which is applicable to all CXR datasets, especially for the Tuberculosis ones.CONCLUSION: Tuberculosis is still one of the most dangerous illnesses in the world that needs vital methods to prevent and detect soon so that patients are able to keep their lives longer. After this research, we are constantly improving the current accuracy of the models and applying the current results of this research for later problems such as detecting the Tuberculosis areas in real-time and supporting doctors to make decisions based on the current status of patients.

Structural textile pattern recognition and processing based on hypergraphs

The humanities, like many other areas of society, are currently undergoing major changes in the wake of digital transformation. However, in order to make collection of digitised material in this area easily accessible, we often still lack adequate search functionality. For instance, digital archives for textiles offer keyword search, which is fairly well understood, and arrange their content following a certain taxonomy, but search functionality at the level of thread structure is still missing. To facilitate the clustering and search, we introduce an approach for recognising similar weaving patterns based on their structures for textile archives. We first represent textile structures using hypergraphs and extract multisets of k-neighbourhoods describing weaving patterns from these graphs. Then, the resulting multisets are clustered using various distance measures and various clustering algorithms (K-Means for simplicity and hierarchical agglomerative algorithms for precision). We evaluate the different variants of our approach experimentally, showing that this can be implemented efficiently (meaning it has linear complexity), and demonstrate its quality to query and cluster datasets containing large textile samples. As, to the best of our knowledge, this is the first practical approach for explicitly modelling complex and irregular weaving patterns usable for retrieval, we aim at establishing a solid baseline.