In today’s digital landscape, where cyber attacks have become the norm, the detection of cyber attacks and threats is critically imperative across diverse domains. Our research presents a new empirical framework for cyber threat modeling, adept at parsing and categorizing cyber-related information from news articles, enhancing real-time vigilance for market stakeholders. At the core of this framework is a fine-tuned BERT model, which we call CANAL - Cyber Activity News Alerting Language Model, tailored for cyber categorization using a novel silver labeling approach powered by Random Forest. We benchmark CANAL against larger, costlier LLMs, including GPT-4, LLaMA, and Zephyr, highlighting their zero to few-shot learning in cyber news classification. CANAL demonstrates superior performance by outperforming all other LLM counterparts in both accuracy and cost-effectiveness. Furthermore, we introduce the Cyber Signal Discovery module, a strategic component designed to efficiently detect emerging cyber signals from news articles. Collectively, CANAL and Cyber Signal Discovery module equip our framework to provide a robust and cost-effective solution for businesses that require agile responses to cyber intelligence.
{"title":"CANAL - Cyber Activity News Alerting Language Model : Empirical Approach vs. Expensive LLMs","authors":"Urjitkumar Patel, Fang-Chun Yeh, Chinmay Gondhalekar","doi":"10.1109/ICAIC60265.2024.10433839","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433839","url":null,"abstract":"In today’s digital landscape, where cyber attacks have become the norm, the detection of cyber attacks and threats is critically imperative across diverse domains. Our research presents a new empirical framework for cyber threat modeling, adept at parsing and categorizing cyber-related information from news articles, enhancing real-time vigilance for market stakeholders. At the core of this framework is a fine-tuned BERT model, which we call CANAL - Cyber Activity News Alerting Language Model, tailored for cyber categorization using a novel silver labeling approach powered by Random Forest. We benchmark CANAL against larger, costlier LLMs, including GPT-4, LLaMA, and Zephyr, highlighting their zero to few-shot learning in cyber news classification. CANAL demonstrates superior performance by outperforming all other LLM counterparts in both accuracy and cost-effectiveness. Furthermore, we introduce the Cyber Signal Discovery module, a strategic component designed to efficiently detect emerging cyber signals from news articles. Collectively, CANAL and Cyber Signal Discovery module equip our framework to provide a robust and cost-effective solution for businesses that require agile responses to cyber intelligence.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"24 6","pages":"1-12"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895351","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433799
T. H. Sardar, Ruhul Amin Hazarika, Bishwajeet Pandey, Guru Prasad M S, Sk Mahmudul Hassan, Radhakrishna Dodmane, Hardik A. Gohel
Objectives: This work aims to develop an automated video summarising methodology and timestamping that uses natural language processing (NLP) tools to extract significant video information.Methods: The methodology comprises extracting the audio from the video, splitting it into chunks by the size of the pauses, and transcribing the audio using Google's speech recognition. The transcribed text is tokenised to create a summary, sentence and word frequencies are calculated, and the most relevant sentences are selected. The summary quality is assessed using ROUGE criteria, and the most important keywords are extracted from the transcript using RAKE.Findings: Our proposed method successfully extracts key points from video lectures and creates text summaries. Timestamping these key points provides valuable context and facilitates navigation within the lecture. Our method combines video-to-text conversion and text summarisation with timestamping key concepts, offering a novel approach to video lecture analysis. Existing video analysis methods focus on keyword extraction or summarisation, while our method offers a more comprehensive approach. Our timestamped key points provide a unique feature compared to other methods. Our method enhances existing video reports by (i) providing concise summaries of key concepts and (ii) enabling quick access to specific information through timestamps. (iii) Facilitating information retrieval through a searchable index. Further research directions: (i) Improve the accuracy of the multi-stage processing pipeline. (ii) Develop techniques to handle diverse accents and pronunciations. (iii) Explore applications of the proposed method to other video genres and types.Application/Improvements: This approach is practical in giving accurate video summaries, saving viewers time and effort when comprehending the main concepts presented in a video.
{"title":"Video key concept extraction using Convolution Neural Network","authors":"T. H. Sardar, Ruhul Amin Hazarika, Bishwajeet Pandey, Guru Prasad M S, Sk Mahmudul Hassan, Radhakrishna Dodmane, Hardik A. Gohel","doi":"10.1109/ICAIC60265.2024.10433799","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433799","url":null,"abstract":"Objectives: This work aims to develop an automated video summarising methodology and timestamping that uses natural language processing (NLP) tools to extract significant video information.Methods: The methodology comprises extracting the audio from the video, splitting it into chunks by the size of the pauses, and transcribing the audio using Google's speech recognition. The transcribed text is tokenised to create a summary, sentence and word frequencies are calculated, and the most relevant sentences are selected. The summary quality is assessed using ROUGE criteria, and the most important keywords are extracted from the transcript using RAKE.Findings: Our proposed method successfully extracts key points from video lectures and creates text summaries. Timestamping these key points provides valuable context and facilitates navigation within the lecture. Our method combines video-to-text conversion and text summarisation with timestamping key concepts, offering a novel approach to video lecture analysis. Existing video analysis methods focus on keyword extraction or summarisation, while our method offers a more comprehensive approach. Our timestamped key points provide a unique feature compared to other methods. Our method enhances existing video reports by (i) providing concise summaries of key concepts and (ii) enabling quick access to specific information through timestamps. (iii) Facilitating information retrieval through a searchable index. Further research directions: (i) Improve the accuracy of the multi-stage processing pipeline. (ii) Develop techniques to handle diverse accents and pronunciations. (iii) Explore applications of the proposed method to other video genres and types.Application/Improvements: This approach is practical in giving accurate video summaries, saving viewers time and effort when comprehending the main concepts presented in a video.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"71 9","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433806
Ade Kurniawan, Y. Ohsita, Masayuki Murata
In multi-sensor systems, certain sensors could have vulnerabilities that may be exploited to produce AEs. However, it is difficult to protect all sensor devices, because the risk of the existence of vulnerable sensor devices increases as the number of sensor devices increases. Therefore, we need a method to protect ML models even if a part of the sensors are compromised by the attacker. One approach is to detect the sensors used by the attacks and remove the detected sensors. However, such reactive defense method has limitations. If some critical sensors that are necessary to distinguish required states are compromised by the attacker, we cannot obtain the suitable output. In this paper, we discuss a strategy to make the system robust against AEs proactively. A system with enough redundancy can work after removing the features from the sensors used in the AEs. That is, we need a metric to check if the system has enough redundancy. In this paper, we define groups of sensors that might be compromised by the same attacker, and we propose a metric called criticality that indicates how important each group of sensors are for classification between two classes. Based on the criticality, we can make the system robust against sensor-based AEs by interactively adding sensors so as to decrease the criticality of any groups of sensors for the classes that must be distinguished.
在多传感器系统中，某些传感器可能存在漏洞，可能会被利用来产生 AE。然而，要保护所有传感器设备是很困难的，因为随着传感器设备数量的增加，存在漏洞的传感器设备的风险也会增加。因此，我们需要一种方法来保护 ML 模型，即使部分传感器被攻击者破坏。一种方法是检测攻击所使用的传感器，并移除检测到的传感器。然而，这种被动防御方法有其局限性。如果一些区分所需状态的关键传感器被攻击者破坏，我们就无法获得合适的输出。在本文中，我们讨论了一种使系统主动抵御 AE 的策略。一个具有足够冗余度的系统可以在去除 AE 所用传感器的特征后正常工作。也就是说，我们需要一个指标来检查系统是否有足够的冗余度。在本文中，我们定义了可能会被同一攻击者入侵的传感器组，并提出了一种称为临界度的指标，它表明了每组传感器对于两个类别之间的分类有多重要。根据临界度，我们可以通过交互式添加传感器来降低任何一组传感器对必须区分的类别的临界度，从而使系统对基于传感器的 AE 具有鲁棒性。
{"title":"Toward robust systems against sensor-based adversarial examples based on the criticalities of sensors.","authors":"Ade Kurniawan, Y. Ohsita, Masayuki Murata","doi":"10.1109/ICAIC60265.2024.10433806","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433806","url":null,"abstract":"In multi-sensor systems, certain sensors could have vulnerabilities that may be exploited to produce AEs. However, it is difficult to protect all sensor devices, because the risk of the existence of vulnerable sensor devices increases as the number of sensor devices increases. Therefore, we need a method to protect ML models even if a part of the sensors are compromised by the attacker. One approach is to detect the sensors used by the attacks and remove the detected sensors. However, such reactive defense method has limitations. If some critical sensors that are necessary to distinguish required states are compromised by the attacker, we cannot obtain the suitable output. In this paper, we discuss a strategy to make the system robust against AEs proactively. A system with enough redundancy can work after removing the features from the sensors used in the AEs. That is, we need a metric to check if the system has enough redundancy. In this paper, we define groups of sensors that might be compromised by the same attacker, and we propose a metric called criticality that indicates how important each group of sensors are for classification between two classes. Based on the criticality, we can make the system robust against sensor-based AEs by interactively adding sensors so as to decrease the criticality of any groups of sensors for the classes that must be distinguished.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"26 5","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433805
Gerard Shu Fuhnwi, Matthew Revelle, Clemente Izurieta
In cybersecurity, Network Intrusion Detection Systems (NIDS) are essential for identifying and preventing malicious activity within computer networks. Machine learning algorithms have been widely applied to NIDS due to their ability to identify complex patterns and anomalies in network traffic. Improvements in the performance of an IDS can be measured by increasing the Matthew Correlation Coefficient (MCC), the reduction of False Alarm Rates (FARs), and the maintenance of up-to-date signatures of the latest attacks to maintain confidentiality, integrity, and availability of services. Integrating machine learning with feature selection for IDSs can help eliminate less important features until the optimal subset of features is achieved, thus improving the NIDS.In this research, we propose an approach for NIDS using XGBoost, a popular gradient boosting algorithm, with Recursive Feature Elimination (RFE) feature selection. We used the NSL-KDD dataset, a benchmark dataset for evaluating NIDS, for training and testing. Our empirical results show that XGBoost with RFE outperforms other popular machine learning algorithms for NIDS on this dataset, achieving the highest MCC for detecting NSL-KDD dataset attacks of type DoS, Probe, U2R, and R2L and very high classification time.
{"title":"Improving Network Intrusion Detection Performance : An Empirical Evaluation Using Extreme Gradient Boosting (XGBoost) with Recursive Feature Elimination","authors":"Gerard Shu Fuhnwi, Matthew Revelle, Clemente Izurieta","doi":"10.1109/ICAIC60265.2024.10433805","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433805","url":null,"abstract":"In cybersecurity, Network Intrusion Detection Systems (NIDS) are essential for identifying and preventing malicious activity within computer networks. Machine learning algorithms have been widely applied to NIDS due to their ability to identify complex patterns and anomalies in network traffic. Improvements in the performance of an IDS can be measured by increasing the Matthew Correlation Coefficient (MCC), the reduction of False Alarm Rates (FARs), and the maintenance of up-to-date signatures of the latest attacks to maintain confidentiality, integrity, and availability of services. Integrating machine learning with feature selection for IDSs can help eliminate less important features until the optimal subset of features is achieved, thus improving the NIDS.In this research, we propose an approach for NIDS using XGBoost, a popular gradient boosting algorithm, with Recursive Feature Elimination (RFE) feature selection. We used the NSL-KDD dataset, a benchmark dataset for evaluating NIDS, for training and testing. Our empirical results show that XGBoost with RFE outperforms other popular machine learning algorithms for NIDS on this dataset, achieving the highest MCC for detecting NSL-KDD dataset attacks of type DoS, Probe, U2R, and R2L and very high classification time.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"259 7","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139896083","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433846
Charlie Grimshaw, Brian Lachine, Taylor Perkins, Emilie Coote
Anomaly detection is a challenge well-suited to machine learning and in the context of information security, the benefits of unsupervised solutions show significant promise. Recent attention to Graph Neural Networks (GNNs) has provided an innovative approach to learn from attributed graphs. Using a GNN encoder-decoder architecture, anomalous edges between nodes can be detected during the reconstruction phase. The aim of this research is to determine whether an unsupervised GNN model can detect anomalous network connections in a static, attributed network. Network logs were collected from four corporate networks and one artificial network using endpoint monitoring tools. A GNN-based anomaly detection system was designed and employed to score and rank anomalous connections between hosts. The model was validated against four realistic experimental scenarios against the four large corporate networks and the smaller artificial network environment. Although quantitative metrics were affected by factors including the scale of the network, qualitative assessments indicated that anomalies from all scenarios were detected. The false positives across each scenario indicate that this model in its current form is useful as an initial triage, though would require further improvement to become a performant detector. This research serves as a promising step for advancing this methodology in detecting anomalous network connections. Future work to improve results includes narrowing the scope of detection to specific threat types and a further focus on feature engineering and selection.
{"title":"Link-based Anomaly Detection with Sysmon and Graph Neural Networks","authors":"Charlie Grimshaw, Brian Lachine, Taylor Perkins, Emilie Coote","doi":"10.1109/ICAIC60265.2024.10433846","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433846","url":null,"abstract":"Anomaly detection is a challenge well-suited to machine learning and in the context of information security, the benefits of unsupervised solutions show significant promise. Recent attention to Graph Neural Networks (GNNs) has provided an innovative approach to learn from attributed graphs. Using a GNN encoder-decoder architecture, anomalous edges between nodes can be detected during the reconstruction phase. The aim of this research is to determine whether an unsupervised GNN model can detect anomalous network connections in a static, attributed network. Network logs were collected from four corporate networks and one artificial network using endpoint monitoring tools. A GNN-based anomaly detection system was designed and employed to score and rank anomalous connections between hosts. The model was validated against four realistic experimental scenarios against the four large corporate networks and the smaller artificial network environment. Although quantitative metrics were affected by factors including the scale of the network, qualitative assessments indicated that anomalies from all scenarios were detected. The false positives across each scenario indicate that this model in its current form is useful as an initial triage, though would require further improvement to become a performant detector. This research serves as a promising step for advancing this methodology in detecting anomalous network connections. Future work to improve results includes narrowing the scope of detection to specific threat types and a further focus on feature engineering and selection.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"14 2","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895373","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433841
Bryan Chung
The following study introduces a novel framework for recognizing plant diseases, tackling the issue of imbalanced datasets, which is prevalent in agriculture, a key sector for many economies. Plant diseases can significantly affect crop quality and yield, making early and accurate detection vital for effective disease management. Traditional Convolutional Neural Networks (CNNs) have shown promise in plant disease recognition but often fall short with non-tomato crops due to class imbalance in datasets. The proposed approach utilizes contrastive learning to train a model on the PlantDoc dataset in a self-supervised manner, allowing it to learn meaningful representations from unlabeled data by maximizing the similarity between images based on disease state rather than species. This method shows a marked improvement in accuracy, achieving 87.42% on the PlantDoc dataset and demonstrating its superiority over existing supervised learning methods. The agnostic nature of the model towards plant species allows for universal application in agriculture, offering a significant tool for disease management and enhancing productivity in both existing farms and future smart farming environments.
{"title":"Addressing Data Imbalance in Plant Disease Recognition through Contrastive Learning","authors":"Bryan Chung","doi":"10.1109/ICAIC60265.2024.10433841","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433841","url":null,"abstract":"The following study introduces a novel framework for recognizing plant diseases, tackling the issue of imbalanced datasets, which is prevalent in agriculture, a key sector for many economies. Plant diseases can significantly affect crop quality and yield, making early and accurate detection vital for effective disease management. Traditional Convolutional Neural Networks (CNNs) have shown promise in plant disease recognition but often fall short with non-tomato crops due to class imbalance in datasets. The proposed approach utilizes contrastive learning to train a model on the PlantDoc dataset in a self-supervised manner, allowing it to learn meaningful representations from unlabeled data by maximizing the similarity between images based on disease state rather than species. This method shows a marked improvement in accuracy, achieving 87.42% on the PlantDoc dataset and demonstrating its superiority over existing supervised learning methods. The agnostic nature of the model towards plant species allows for universal application in agriculture, offering a significant tool for disease management and enhancing productivity in both existing farms and future smart farming environments.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"9 4","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895498","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433804
Trishla Shah, Raghav V. Sampangi, Angela Siegel
Over the years, mobile applications have brought about transformative changes in user interactions with digital services. Many of these apps however, are free and offer convenience at the cost of exchanging personal data. This convenience, however, comes with inherent risks to user privacy and security. This paper introduces a comprehensive methodology that evaluates the risks associated with sharing sensitive data through mobile applications. Building upon the Hierarchical Weighted Risk Scoring Model (HWRSM), this paper proposes an evaluation methodology for HWRSM, keeping in mind the implications of such risk scoring on real-world security scenarios. The methodology employs innovative risk scoring, considering various factors to assess potential security vulnerabilities related to sensitive terms. Practical assessments involving diverse set of Android applications, particularly in data-intensive categories, reveal insights into data privacy practices, vulnerabilities, and alignment with HWRSM scores. By offering insights into testing, validation, real-world findings, and model effectiveness, the paper aims to provide practical considerations to mobile application security discussions, facilitating informed approaches to address security and privacy concerns.
{"title":"Risk-Aware Mobile App Security Testing: Safeguarding Sensitive User Inputs","authors":"Trishla Shah, Raghav V. Sampangi, Angela Siegel","doi":"10.1109/ICAIC60265.2024.10433804","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433804","url":null,"abstract":"Over the years, mobile applications have brought about transformative changes in user interactions with digital services. Many of these apps however, are free and offer convenience at the cost of exchanging personal data. This convenience, however, comes with inherent risks to user privacy and security. This paper introduces a comprehensive methodology that evaluates the risks associated with sharing sensitive data through mobile applications. Building upon the Hierarchical Weighted Risk Scoring Model (HWRSM), this paper proposes an evaluation methodology for HWRSM, keeping in mind the implications of such risk scoring on real-world security scenarios. The methodology employs innovative risk scoring, considering various factors to assess potential security vulnerabilities related to sensitive terms. Practical assessments involving diverse set of Android applications, particularly in data-intensive categories, reveal insights into data privacy practices, vulnerabilities, and alignment with HWRSM scores. By offering insights into testing, validation, real-world findings, and model effectiveness, the paper aims to provide practical considerations to mobile application security discussions, facilitating informed approaches to address security and privacy concerns.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"262 6","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139896081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433844
Ahamadali Jamali, Shahin Alipour, Audrey Rah
Auto-labeling of text is a useful and necessary technique for creating large and high-quality training data sets for machine learning models. Label-free sentiment classification is a challenging semi-supervised task in the natural language processing domain. This study leveraged the weak supervision framework to generate weak labels in three categories for millions of news headlines from Australian Broadcasting Corporation (ABC). A Bidirectional Gate Recurrent Unit (BiGRU) was then trained with neural network dense layers to achieve a validation accuracy of 96.76% with 99.99% accuracy. The performance of this method was also compared with traditional and deep learning natural language processing techniques.
{"title":"Leveraging Weak Supervision and BiGRU Neural Networks for Sentiment Analysis on Label-Free News Headlines","authors":"Ahamadali Jamali, Shahin Alipour, Audrey Rah","doi":"10.1109/ICAIC60265.2024.10433844","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433844","url":null,"abstract":"Auto-labeling of text is a useful and necessary technique for creating large and high-quality training data sets for machine learning models. Label-free sentiment classification is a challenging semi-supervised task in the natural language processing domain. This study leveraged the weak supervision framework to generate weak labels in three categories for millions of news headlines from Australian Broadcasting Corporation (ABC). A Bidirectional Gate Recurrent Unit (BiGRU) was then trained with neural network dense layers to achieve a validation accuracy of 96.76% with 99.99% accuracy. The performance of this method was also compared with traditional and deep learning natural language processing techniques.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"3 3","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139895509","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433847
C. Mutongi, Billy Rigava
The stone age did not end because there were no more stones, it ended because of continuous improvement, innovation, creativity and learning. Local government has always been around since time immemorial. Even in the Stone Age period there was some form of local government, leaning and continuous improvement. In this DVUCADD environment, an environment characterized by dynamic, volatile, uncertain, ambiguous, diversity and disruptive phenomena, cities should be in a position to employ Peter Senge’s fifth discipline in order to survive and be in a position to learn faster. The Local government in Africa and Zimbabwe in particular has the role of proving a range of vital learning city services delivery for residents and organisations in defined areas. Among them are well known functions such as social services like primary education, libraries, vocational training and recreational facilities. Local government administration has a great role to play in bringing citizenry’s lifelong learning, engagement and participation. This then brings in economic and social development. One of the important aspects that ever happened in our life, is when Peter Senge came up with the fifth discipline that any organisation can apply in order to achieve a learning organisation. These disciplines are personal mastery, mental models, shared vision, team learning and systems thinking. The City of Harare is used as a case study in the application of Peter Senge’s fifth discipline to foster the learning city concept.
{"title":"The Application of the Fifth Discipline Strategies in the Learning City Concept","authors":"C. Mutongi, Billy Rigava","doi":"10.1109/ICAIC60265.2024.10433847","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433847","url":null,"abstract":"The stone age did not end because there were no more stones, it ended because of continuous improvement, innovation, creativity and learning. Local government has always been around since time immemorial. Even in the Stone Age period there was some form of local government, leaning and continuous improvement. In this DVUCADD environment, an environment characterized by dynamic, volatile, uncertain, ambiguous, diversity and disruptive phenomena, cities should be in a position to employ Peter Senge’s fifth discipline in order to survive and be in a position to learn faster. The Local government in Africa and Zimbabwe in particular has the role of proving a range of vital learning city services delivery for residents and organisations in defined areas. Among them are well known functions such as social services like primary education, libraries, vocational training and recreational facilities. Local government administration has a great role to play in bringing citizenry’s lifelong learning, engagement and participation. This then brings in economic and social development. One of the important aspects that ever happened in our life, is when Peter Senge came up with the fifth discipline that any organisation can apply in order to achieve a learning organisation. These disciplines are personal mastery, mental models, shared vision, team learning and systems thinking. The City of Harare is used as a case study in the application of Peter Senge’s fifth discipline to foster the learning city concept.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"95 2","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139896101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-07DOI: 10.1109/ICAIC60265.2024.10433802
Rathinaraja Jeyaraj, B. Subramanian, Karnam Yogesh, Aobo Jin, Hardik A. Gohel
Besides biometrics, face authentication is quite popular on smart devices like smartphones and other electronic gadgets to verify and authenticate individuals. In the face authentication method, there is a chance of spoofing attacks, in which a static image or recorded video can be substituted for a real person’s face to breach security and gain access. To solve this problem, smart devices use additional hardware like a dual camera or an infrared sensor, which adds extra cost, weight, and incompatibility to different gadgets. Alternatively, software-based methods may be confused with a video of the user to gain the access. To overcome these problems, in this paper, we present a framework, YSAF, that combines Yolo v8 object detection, spatial attention, and fast Fourier transform (FFT) to restrict facial-based spoofing attacks without additional hardware. In YSAF, spatial attention is first used to focus on relevant features and reduce noise in the input image. Next, frequency analysis through FFT is applied to embed information in the collected images to help the classification model differentiate live faces from static ones. As a final step, Yolo detects whether the object present in the collected images is real or fake (spoof). The YSAF is trained using real images collected by volunteers from different sources and pre-processed with spatial attention and FFT before training with Yolo. The results show that the YSAF accurately blocks spoofing attacks with still images/videos in real-time.
{"title":"YSAF: Yolo with Spatial Attention and FFT to Detect Face Spoofing Attacks","authors":"Rathinaraja Jeyaraj, B. Subramanian, Karnam Yogesh, Aobo Jin, Hardik A. Gohel","doi":"10.1109/ICAIC60265.2024.10433802","DOIUrl":"https://doi.org/10.1109/ICAIC60265.2024.10433802","url":null,"abstract":"Besides biometrics, face authentication is quite popular on smart devices like smartphones and other electronic gadgets to verify and authenticate individuals. In the face authentication method, there is a chance of spoofing attacks, in which a static image or recorded video can be substituted for a real person’s face to breach security and gain access. To solve this problem, smart devices use additional hardware like a dual camera or an infrared sensor, which adds extra cost, weight, and incompatibility to different gadgets. Alternatively, software-based methods may be confused with a video of the user to gain the access. To overcome these problems, in this paper, we present a framework, YSAF, that combines Yolo v8 object detection, spatial attention, and fast Fourier transform (FFT) to restrict facial-based spoofing attacks without additional hardware. In YSAF, spatial attention is first used to focus on relevant features and reduce noise in the input image. Next, frequency analysis through FFT is applied to embed information in the collected images to help the classification model differentiate live faces from static ones. As a final step, Yolo detects whether the object present in the collected images is real or fake (spoof). The YSAF is trained using real images collected by volunteers from different sources and pre-processed with spatial attention and FFT before training with Yolo. The results show that the YSAF accurately blocks spoofing attacks with still images/videos in real-time.","PeriodicalId":517265,"journal":{"name":"2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC)","volume":"258 6","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2024-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139896085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：