The protection of Industrial Control Systems (ICS) that are employed in public critical infrastructures is of utmost importance due to catastrophic physical damages cyberattacks may cause. The research community requires testbeds for validation and comparing various intrusion detection algorithms to protect ICS. However, there exist high barriers to entry for research and education in the ICS cybersecurity domain due to expensive hardware, software, and inherent dangers of manipulating real-world systems. To close the gap, built upon recently developed 3D high-fidelity simulators, we further showcase our integrated framework to automatically launch cyberattacks, collect data, train machine learning models, and evaluate for practical chemical and manufacturing processes. On our testbed, we validate our proposed intrusion detection model called Minimal Threshold and Window SVM (MinTWin SVM) that utilizes unsupervised machine learning via a one-class SVM in combination with a sliding window and classification threshold. Results show that MinTWin SVM minimizes false positives and is responsive to physical process anomalies. Furthermore, we incorporate our framework with ICS cybersecurity education by using our dataset in an undergraduate machine learning course where students gain hands-on experience in practicing machine learning theory with a practical ICS dataset. All of our implementations have been open-sourced.
Web application security poses ongoing challenges for organizations, and researchers have increasingly turned to machine and deep learning techniques to address vulnerabilities such as SQL injection and Cross Site Scripting. While these studies have made notable progress, the vulnerability to adversarial attacks remains a significant challenge. In this work, we aim to explore the impact of adversarial examples on machine and deep learning applications for detecting web vulnerabilities. We propose an evolutionary algorithm to create effective adversarial samples for bypassing various detection systems by iterative exploration of the search space. The effectiveness of our proposed approach is validated through rigorous testing on different diverse SQL injection detection systems.
The COVID-19 pandemic has had a global impact on communities, economies, and healthcare systems. To control the virus's spread, numerous epidemiological investigations have been made available online, leading to a growing demand for automated tools to extract valuable information from case reports and reduce the burden on news reporters. In response to this growing need, we have meticulously curated a comprehensive data set of COVID-19 epidemiological investigation corpora, specifically designed for named entity recognition (NER) applications. This data set enables researchers and analysts to efficiently identify and extract key information from the case reports, streamlining the process of understanding and communicating the findings. To further enhance the effectiveness of NER in the context of epidemiological investigations, we evaluated and compared the performance of three cutting-edge, pre-trained model-based methods: BERT-BiLSTM-CRF, ERNIE-BiLSTM- CRF and ALBERT-BiLSTM-CRF. All techniques demonstrated impressive performance in recognizing named entities within the case reports, showcasing their potential to revolutionize the way in which epidemiological data is analyzed and disseminated. By leveraging these advanced NER techniques, we aim to facilitate more accurate and timely reporting, ultimately contributing to better-informed decision-making processes and improved public health outcomes.
Intelligent transportation systems have become increasingly important for efficient traffic management and road safety. Vehicle classification is a fundamental task in these systems, enabling various applications such as traffic monitoring, congestion management, and accident prevention. Traditional methods for vehicle classification heavily rely on visual or sensor-based data, such as images or radar signals. However, these methods may encounter limitations in adverse weather conditions, poor lighting, or occlusion scenarios. To address these limitations, this paper introduces a novel approach for vehicle classification using seismic data, which captures the vibrations generated by vehicles and is less susceptible to environmental factors. The proposed approach leverages the fractional wavelet domain to extract both time and frequency signatures of vehicles from the seismic data effectively. These signatures are then used to train a recurrent neural network that captures the temporal features of the input, thereby facilitating vehicle classification. Additionally, the proposed approach incorporates various modules to improve the generalization and robustness of the deep model against noise. The implementation of the proposed approach on realistic data demonstrated its ability to classify vehicles with an accuracy of 98%, improving upon the state-of-the-art techniques.
Accurately predicting hotspot reprint paragraphs can timely provide valuable clues for topic selection, thereby improving the influence of the disseminated content. Most existing works in media reprint analysis focus on mining reprint relationships and reprint patterns. Meanwhile, few works predict the hotspot reprint paragraph from a fine-grained level. The writing style reflects the structure and semantic logic of the article to some extent. Thus, the challenge is to determine how to effectively incorporate writing style features into the semantic analysis while also reasoning deeply about the semantic relevance between sections of the article. This paper proposes a multi-perspective relevance collaborative modeling method called MPRCM-TS. It integrates writing styles of titles into the semantic representations and deeply mines the multi-perspective semantic relevance between the title and paragraphs on the basis of the attention mechanism. Simultaneously, multiple loss functions collaborate to enhance the parameter optimization ability. We evaluate the performance of the proposed model on a real-world dataset, and the experimental results demonstrate the efficacy.
