Face Detection plays a key role in “always-on” applications such as mobile phone unlock or smart doorbells. Deep learning-based face detection solutions have demonstrated state-of-art performance in terms of accuracy; however generally, the improved accuracy comes with a large computation and memory requirement overhead. This can result in high energy consumption which is a significant cost that can overrun the energy budget especially in battery powered systems. Recent solutions to this problem have advocated the use of a low power always-on sensor running a rudimentary algorithm that can merely indicate the ‘presence’ of a face with low accuracy and in turn ‘wake-up’ a more powerful device executing a high accuracy face detection algorithm. In this paper we present the design of two deeply quantized (binarized) light weight face presence detection deep learning based models that can function as wake up models. The models achieve high accuracy> 98% with a corresponding memory footprint being limited between 3KB and 100KB allowing them to be deployed in highly resource constrained ‘always-on’ embedded platforms.
{"title":"Tiny Face Presence Detector using Hybrid Binary Neural Network","authors":"Manav Chandna, Pratishtha Bhatia, Surinder-pal Singh, Saumya Suneja","doi":"10.1109/ICITIIT57246.2023.10068573","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068573","url":null,"abstract":"Face Detection plays a key role in “always-on” applications such as mobile phone unlock or smart doorbells. Deep learning-based face detection solutions have demonstrated state-of-art performance in terms of accuracy; however generally, the improved accuracy comes with a large computation and memory requirement overhead. This can result in high energy consumption which is a significant cost that can overrun the energy budget especially in battery powered systems. Recent solutions to this problem have advocated the use of a low power always-on sensor running a rudimentary algorithm that can merely indicate the ‘presence’ of a face with low accuracy and in turn ‘wake-up’ a more powerful device executing a high accuracy face detection algorithm. In this paper we present the design of two deeply quantized (binarized) light weight face presence detection deep learning based models that can function as wake up models. The models achieve high accuracy> 98% with a corresponding memory footprint being limited between 3KB and 100KB allowing them to be deployed in highly resource constrained ‘always-on’ embedded platforms.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126424571","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068571
Fredy Jingga, Zara Handra Fitria, Julian Alfi, Andreano Dwi Kusumaiati
The development of music streaming is rapidly increasing on a global scale. The official website of the International Federation of Phonographic Industry (IFPI) states that on a global scale, in 2020, 62.1% of the music industry's revenue is due to streaming services. This study's objective is to investigate the elements that have an impact on Indonesians' use of streaming music. This research initially wanted to explore the link between advantages and convenience and consumers' behavioural intention to utilize streaming music. Using data from 117 respondents obtained from social media with the Structural Equation Modeling (PLS-SEM) method as a data processor. This research shows that benefits, features, habits, and ease of use are behavioural intentions of music streaming users in Indonesia. The findings of this study will eventually be used to advise music streaming platform providers about critical aspects affecting music streaming customers in Indonesia.
{"title":"Factors influenced user in Using Streaming Music Applications Using the TAM Method: Technology Acceptance Model","authors":"Fredy Jingga, Zara Handra Fitria, Julian Alfi, Andreano Dwi Kusumaiati","doi":"10.1109/ICITIIT57246.2023.10068571","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068571","url":null,"abstract":"The development of music streaming is rapidly increasing on a global scale. The official website of the International Federation of Phonographic Industry (IFPI) states that on a global scale, in 2020, 62.1% of the music industry's revenue is due to streaming services. This study's objective is to investigate the elements that have an impact on Indonesians' use of streaming music. This research initially wanted to explore the link between advantages and convenience and consumers' behavioural intention to utilize streaming music. Using data from 117 respondents obtained from social media with the Structural Equation Modeling (PLS-SEM) method as a data processor. This research shows that benefits, features, habits, and ease of use are behavioural intentions of music streaming users in Indonesia. The findings of this study will eventually be used to advise music streaming platform providers about critical aspects affecting music streaming customers in Indonesia.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"4 3","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120900056","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068638
M. Vivekanandan, Praveen Kumar Premkamal, C. I. Johnpaul, Silambarasan Elkana Ebinazer
Blockchain based distributed ledger mechanism has got a wide range of applications in this era. The degree of security measurement is always a bottleneck. Since there are technologies to break it. Data sharing through cloud for smart cities, collaborative actions, remote activities based on the data at the source, etc., need to be secure and free from masquerading and tampering. In most of the cases the data is pushed into the cloud from access points, sensors, or remote access centers. Preventing the data access and identifying anonymous access to these sensors require an enhanced security mechanism that prevents the inconsistent data to be transferred to the cloud. We propose a blockchain based enhanced security system that protects the data from the access point it leaves for the cloud using a distributed ledger. The consensus mechanism ensures the trust of existing sources during the data transfer from the source to the cloud. The trust generated by the subsequent data blocks with the security hash key ensure the integrity of the data and validity of the actual source. This prevent the illegal access to the data sharing points. We have verified the degree of security offered by our proposed model using informal analysis. We found that our method has improved the security of data access.
{"title":"Blockchain based Secure Data Storage Verification Algorithm for Smart City Environment","authors":"M. Vivekanandan, Praveen Kumar Premkamal, C. I. Johnpaul, Silambarasan Elkana Ebinazer","doi":"10.1109/ICITIIT57246.2023.10068638","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068638","url":null,"abstract":"Blockchain based distributed ledger mechanism has got a wide range of applications in this era. The degree of security measurement is always a bottleneck. Since there are technologies to break it. Data sharing through cloud for smart cities, collaborative actions, remote activities based on the data at the source, etc., need to be secure and free from masquerading and tampering. In most of the cases the data is pushed into the cloud from access points, sensors, or remote access centers. Preventing the data access and identifying anonymous access to these sensors require an enhanced security mechanism that prevents the inconsistent data to be transferred to the cloud. We propose a blockchain based enhanced security system that protects the data from the access point it leaves for the cloud using a distributed ledger. The consensus mechanism ensures the trust of existing sources during the data transfer from the source to the cloud. The trust generated by the subsequent data blocks with the security hash key ensure the integrity of the data and validity of the actual source. This prevent the illegal access to the data sharing points. We have verified the degree of security offered by our proposed model using informal analysis. We found that our method has improved the security of data access.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116070664","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068577
Ashish Bajaj, D. Vishwakarma
News websites need to divide their articles into categories that make it easier for readers to find news of their interest. Recent deep-learning models have excelled in this news classification task. Despite the tremendous success of deep learning models in NLP-related tasks, it is vulnerable to adversarial attacks, which lead to misclassification of the news category. An adversarial text is generated by changing a few words or characters in a way that retains the overall semantic similarity of news for a human reader but deceives the machine into giving inaccurate predictions. This paper presents the vulnerability in news classification by generating adversarial text using various state-of-the-art attack algorithms. We have compared and analyzed the behavior of different models, including the powerful transformer model, BERT, and the widely used Word-CNN and LSTM models trained on AG news classification dataset. We have evaluated the potential results by calculating Attack Success Rates (ASR) for each model. The results show that it is possible to automatically bypass News topic classification mechanisms, resulting in repercussions for current policy measures.
{"title":"Exposing the Vulnerabilities of Deep Learning Models in News Classification","authors":"Ashish Bajaj, D. Vishwakarma","doi":"10.1109/ICITIIT57246.2023.10068577","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068577","url":null,"abstract":"News websites need to divide their articles into categories that make it easier for readers to find news of their interest. Recent deep-learning models have excelled in this news classification task. Despite the tremendous success of deep learning models in NLP-related tasks, it is vulnerable to adversarial attacks, which lead to misclassification of the news category. An adversarial text is generated by changing a few words or characters in a way that retains the overall semantic similarity of news for a human reader but deceives the machine into giving inaccurate predictions. This paper presents the vulnerability in news classification by generating adversarial text using various state-of-the-art attack algorithms. We have compared and analyzed the behavior of different models, including the powerful transformer model, BERT, and the widely used Word-CNN and LSTM models trained on AG news classification dataset. We have evaluated the potential results by calculating Attack Success Rates (ASR) for each model. The results show that it is possible to automatically bypass News topic classification mechanisms, resulting in repercussions for current policy measures.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127688298","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068685
Adarsh Vernekar, Akash Sakhare, Prashant Bhapkar, S. Jadhav, Rahul B. Adhao
Every industry is expanding too quickly and adjusting to this new technology as it develops. Since it has the potential to deliver more precise and economical patient care, healthcare data management has recently attracted a lot of attention. Even today, many hospitals hold their own autonomous record management system, which causes security issues. In centralized record management systems, data privacy, centralized data stewardship, and system vulnerability problems affect traditional client-server-based and cloud-based health data management systems. Blockchain technology has a promising future in the healthcare industry because of its immutability, transparency, privacy, and security properties, which can address certain critical problems with the health management system. A more patient-oriented approach in healthcare systems is required to improve the accuracy and transparency of medical data. In healthcare systems, health records are the most sensitive asset that must be unique and protected across the system. Our objective is to showcase the potential use of blockchain technology in health record management systems in hospitals. In this paper, we demonstrate a health record management system that uses blockchain technology to store the medical records of a patient across multiple hospitals. The proposed system will mainly help in maintaining consistency issues related to data along with improved security in the system.
{"title":"Blockchain Based Record Management System in Hospitals","authors":"Adarsh Vernekar, Akash Sakhare, Prashant Bhapkar, S. Jadhav, Rahul B. Adhao","doi":"10.1109/ICITIIT57246.2023.10068685","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068685","url":null,"abstract":"Every industry is expanding too quickly and adjusting to this new technology as it develops. Since it has the potential to deliver more precise and economical patient care, healthcare data management has recently attracted a lot of attention. Even today, many hospitals hold their own autonomous record management system, which causes security issues. In centralized record management systems, data privacy, centralized data stewardship, and system vulnerability problems affect traditional client-server-based and cloud-based health data management systems. Blockchain technology has a promising future in the healthcare industry because of its immutability, transparency, privacy, and security properties, which can address certain critical problems with the health management system. A more patient-oriented approach in healthcare systems is required to improve the accuracy and transparency of medical data. In healthcare systems, health records are the most sensitive asset that must be unique and protected across the system. Our objective is to showcase the potential use of blockchain technology in health record management systems in hospitals. In this paper, we demonstrate a health record management system that uses blockchain technology to store the medical records of a patient across multiple hospitals. The proposed system will mainly help in maintaining consistency issues related to data along with improved security in the system.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132808859","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068616
M. Ezhilarasi, Anand Kumar, M. Shanmugapriya, Anshul Ghanshala, Anshika Gupta
To minimize overall healthcare costs and enhance workflows and processes, remote health monitoring solutions are needed in both clinics and at home. One of the most effective communication technologies, the Internet of Things (IoT) offers the ability of integrated data access and fusion across a variety of applications. Depending on how each person's role is defined, users and qualified professionals (like doctors and nurses in the medical industry) may be able to access data. The goal of the Internet of Things in the healthcare industry is to redefine the healthcare system by bringing together all involved authorities and cutting-edge technology makes the most of the data shared between intimately linked technologies that use the IoT platform. IoT is generally anticipated to provide an enhanced device, scheme, and application connectivity that extend over machine-to-machine communications. In the past few years, the development of wearable sensors has offered ease, simplicity, and better health. By making medical sensors smaller and less expensive, technological advancements have boosted the use of these devices. The expertise and abilities of healthcare services, such as remote health monitoring, surgery, rehabilitation, and therapy are improved through medical sensors. Fog computing techniques are also added to enhance precision medicine, obtain real-time data processing, and prevent the connection from failing. As a result, the operating environment for devices is more nimble and local. In this regard, this study suggests an architecture model for the healthcare domain that incorporates the technologies of body area networks, IoT, and Fog computing. The key contribution is to boost the capabilities and resources of IoT devices by using an intermediate Fog layer close to the edge to get beyond IoT restrictions. Experiments show that when compared to other standard architecture, the suggested model can reach a 75% faster response time. The evaluation's results supported the suggested model's ability to accomplish its objectives while maintaining application performance.
{"title":"Integrated Healthcare Monitoring System using Wireless Body Area Networks and Internet of Things","authors":"M. Ezhilarasi, Anand Kumar, M. Shanmugapriya, Anshul Ghanshala, Anshika Gupta","doi":"10.1109/ICITIIT57246.2023.10068616","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068616","url":null,"abstract":"To minimize overall healthcare costs and enhance workflows and processes, remote health monitoring solutions are needed in both clinics and at home. One of the most effective communication technologies, the Internet of Things (IoT) offers the ability of integrated data access and fusion across a variety of applications. Depending on how each person's role is defined, users and qualified professionals (like doctors and nurses in the medical industry) may be able to access data. The goal of the Internet of Things in the healthcare industry is to redefine the healthcare system by bringing together all involved authorities and cutting-edge technology makes the most of the data shared between intimately linked technologies that use the IoT platform. IoT is generally anticipated to provide an enhanced device, scheme, and application connectivity that extend over machine-to-machine communications. In the past few years, the development of wearable sensors has offered ease, simplicity, and better health. By making medical sensors smaller and less expensive, technological advancements have boosted the use of these devices. The expertise and abilities of healthcare services, such as remote health monitoring, surgery, rehabilitation, and therapy are improved through medical sensors. Fog computing techniques are also added to enhance precision medicine, obtain real-time data processing, and prevent the connection from failing. As a result, the operating environment for devices is more nimble and local. In this regard, this study suggests an architecture model for the healthcare domain that incorporates the technologies of body area networks, IoT, and Fog computing. The key contribution is to boost the capabilities and resources of IoT devices by using an intermediate Fog layer close to the edge to get beyond IoT restrictions. Experiments show that when compared to other standard architecture, the suggested model can reach a 75% faster response time. The evaluation's results supported the suggested model's ability to accomplish its objectives while maintaining application performance.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"126 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133906551","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068637
Kalit Naresh Inani, K. S. Sangwan, Dhiraj
The recent advancement in the state of art technologies for drones and their reduced cost have made them highly accessible to the general public. Though their application is increasing in several domains, they raise security and privacy issues for military bases and civilians. To prevent this, drone detection and identification using RF signals is explored. The dataset considered in this experimental study is DroneRF dataset. Initially, the raw RF data is preprocessed to extract most relevant features using power spectral density technique which are further utilized for training machine learning classifiers such as XGBoost which gave the best accuracy for 2,4 and 10 category. The XGBoost algorithm with PSD features provides 100%, 100%, and 99.73% accuracy for 2, 4 and 10 category based data. To explore the possibility of feature fusion, another experiment was done XGBoost gave 99.13%, 99.11%, and 93.84% accuracy for 2,4 and 10 class problem. To investigate the usage of deep learning techniques, 1DCNN was used which provides 100%, 94.31%, and 86.29% accuracy scores. The final experiment was done using a Hybrid approach where 1DCNN based feature extractor and XGBoost classifier provides 100%, 99.82%, and 99.51% accuracies.
{"title":"Machine Learning based framework for Drone Detection and Identification using RF signals","authors":"Kalit Naresh Inani, K. S. Sangwan, Dhiraj","doi":"10.1109/ICITIIT57246.2023.10068637","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068637","url":null,"abstract":"The recent advancement in the state of art technologies for drones and their reduced cost have made them highly accessible to the general public. Though their application is increasing in several domains, they raise security and privacy issues for military bases and civilians. To prevent this, drone detection and identification using RF signals is explored. The dataset considered in this experimental study is DroneRF dataset. Initially, the raw RF data is preprocessed to extract most relevant features using power spectral density technique which are further utilized for training machine learning classifiers such as XGBoost which gave the best accuracy for 2,4 and 10 category. The XGBoost algorithm with PSD features provides 100%, 100%, and 99.73% accuracy for 2, 4 and 10 category based data. To explore the possibility of feature fusion, another experiment was done XGBoost gave 99.13%, 99.11%, and 93.84% accuracy for 2,4 and 10 class problem. To investigate the usage of deep learning techniques, 1DCNN was used which provides 100%, 94.31%, and 86.29% accuracy scores. The final experiment was done using a Hybrid approach where 1DCNN based feature extractor and XGBoost classifier provides 100%, 99.82%, and 99.51% accuracies.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130039831","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068666
Ruchitaa Raj N R, Nandhakumar Raj S, V. M
Web scraping can be done using many languages such as C++, Java, JavaScript, PhP, Python, Ruby, etc. Among them, Python stands to be the most powerful language with lots of inbuilt libraries that supports web scraping, extensive support for third-party open-source libraries, and higher speeds compared to other languages. Python libraries for web scraping are designed for fast and highly accurate data extraction. There are many libraries available for web scraping and the developer can choose the respective library in accordance with their scraping application. This paper focuses on the study of several web scraping tools and techniques and analyze the performance of those tools and present the statistical significance of the results.
{"title":"Web Scrapping Tools and Techniques: A Brief Survey","authors":"Ruchitaa Raj N R, Nandhakumar Raj S, V. M","doi":"10.1109/ICITIIT57246.2023.10068666","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068666","url":null,"abstract":"Web scraping can be done using many languages such as C++, Java, JavaScript, PhP, Python, Ruby, etc. Among them, Python stands to be the most powerful language with lots of inbuilt libraries that supports web scraping, extensive support for third-party open-source libraries, and higher speeds compared to other languages. Python libraries for web scraping are designed for fast and highly accurate data extraction. There are many libraries available for web scraping and the developer can choose the respective library in accordance with their scraping application. This paper focuses on the study of several web scraping tools and techniques and analyze the performance of those tools and present the statistical significance of the results.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130787149","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068722
Rithvik Senthil, Lakshana Ravishankar, Snofy D. Dunston, M. V
In this study, we analyse the impact of the Universal Adversarial Perturbation Attack on the Inception-ResNet-v1 model using the lung CT scan dataset for COVID-19 classification and the retinal OCT scan dataset for Diabetic Macular Edema (DME) classification. The effectiveness of adversarial retraining as a suitable defense mechanism against this attack is examined. This study is categorised into three sections - the implementation of the Inception-ResNet-v1 model, the effect of the attack and the adversarial retraining.
{"title":"Universal Adversarial Perturbation Attack on the Inception-Resnet-v1 model and the Effectiveness of Adversarial Retraining as a Suitable Defense Mechanism","authors":"Rithvik Senthil, Lakshana Ravishankar, Snofy D. Dunston, M. V","doi":"10.1109/ICITIIT57246.2023.10068722","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068722","url":null,"abstract":"In this study, we analyse the impact of the Universal Adversarial Perturbation Attack on the Inception-ResNet-v1 model using the lung CT scan dataset for COVID-19 classification and the retinal OCT scan dataset for Diabetic Macular Edema (DME) classification. The effectiveness of adversarial retraining as a suitable defense mechanism against this attack is examined. This study is categorised into three sections - the implementation of the Inception-ResNet-v1 model, the effect of the attack and the adversarial retraining.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132107635","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 : 2023-02-11DOI: 10.1109/ICITIIT57246.2023.10068588
Sukesh Babu V S, Rahul Raman
Knowledge of pedestrian's walking direction is very crucial in multiple domains of video processing. This paper proposes a graph based, robust and light weighted model for direction estimation of pedestrian's walk by using the property of perspective distortion. Here perspective distortion pattern is used as an advantage in estimation of direction. The graph-based solution uses 3 parallel approaches for estimating the direction: Perspective Distortion Graph, Centroid Displacement and Clustering of Vanishing point. A pedestrian in a frame can be identified by bounding boxes. The temporal dimensional features of bounding boxes are height and width and these features changes for a particular object from frame to frame as the objects moves. These changes are unique for each direction for each object. These changes in dimension along with clustering of vanishing point and centroid displacement is used for the assesment of the pedestrian's walk direction. All the existing approaches need some sort of pre-processing on the frames, which makes the model more complex and time consuming. In the proposed model, the video sequence is applied on YOLO V4 algorithm and bounding boxes are obtained. By analysing the changes from frame to frame for the dimensions, graphs are plotted and minimum and maximum extremas are detected form the graph by eliminating soft extremas. After that envelope is placed for the graph and an average line is drawn based on the envelope, which will give the inference about the direction of walk of the pedestrian. The perspective distortion graph will not give accurate estimation for all directions. So, Centroid displacement and clustering of vanishing point are also used for direction estimation. The result obtained from the three methods are combined and form a robust model. For accurately estimating walk direction, the movement is limited to 8 different directions. For experiment, NITR Conscious Walk dataset and self-acquired dataset are used. With balanced accuracy of 97.003% and 96.25% and a false positive rate of 0.63% and 0.65%, respectively, the model produces good results for the above dataset.
{"title":"Pedestrian Direction Estimation: An Approach via Perspective Distortion Patterns","authors":"Sukesh Babu V S, Rahul Raman","doi":"10.1109/ICITIIT57246.2023.10068588","DOIUrl":"https://doi.org/10.1109/ICITIIT57246.2023.10068588","url":null,"abstract":"Knowledge of pedestrian's walking direction is very crucial in multiple domains of video processing. This paper proposes a graph based, robust and light weighted model for direction estimation of pedestrian's walk by using the property of perspective distortion. Here perspective distortion pattern is used as an advantage in estimation of direction. The graph-based solution uses 3 parallel approaches for estimating the direction: Perspective Distortion Graph, Centroid Displacement and Clustering of Vanishing point. A pedestrian in a frame can be identified by bounding boxes. The temporal dimensional features of bounding boxes are height and width and these features changes for a particular object from frame to frame as the objects moves. These changes are unique for each direction for each object. These changes in dimension along with clustering of vanishing point and centroid displacement is used for the assesment of the pedestrian's walk direction. All the existing approaches need some sort of pre-processing on the frames, which makes the model more complex and time consuming. In the proposed model, the video sequence is applied on YOLO V4 algorithm and bounding boxes are obtained. By analysing the changes from frame to frame for the dimensions, graphs are plotted and minimum and maximum extremas are detected form the graph by eliminating soft extremas. After that envelope is placed for the graph and an average line is drawn based on the envelope, which will give the inference about the direction of walk of the pedestrian. The perspective distortion graph will not give accurate estimation for all directions. So, Centroid displacement and clustering of vanishing point are also used for direction estimation. The result obtained from the three methods are combined and form a robust model. For accurately estimating walk direction, the movement is limited to 8 different directions. For experiment, NITR Conscious Walk dataset and self-acquired dataset are used. With balanced accuracy of 97.003% and 96.25% and a false positive rate of 0.63% and 0.65%, respectively, the model produces good results for the above dataset.","PeriodicalId":170485,"journal":{"name":"2023 4th International Conference on Innovative Trends in Information Technology (ICITIIT)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133796798","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：