Pub Date : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298118
N. Muhammed, Nour Ali, K. Salah, Ayub Khan
RTL verification is still one the most challenging activities in digital system development as it is still the bottleneck in the time-to-market for an integrated circuit development cycle. Thus reducing verification time is one of the most important targets. In this paper, a tool is developed to generate automatic tests from SystemVerilog assertions or SystemVerilog Coverage. The proposed tool is tested using different memory modules starting from single port RAM through Multiple ports RAM, FIFO and the DDRx families. The performance, regarding the runtime, has been compared with the handcrafted test case generation process. Moreover, the performance has been compared with other automatic test generation tools. Results shows the effectiveness of the proposed design. The proposed tool excelled in terms of its run-time, complexity, and coverage percentage.
{"title":"Assertion and Coverage Driven Test Generation Tool for RTL Designs","authors":"N. Muhammed, Nour Ali, K. Salah, Ayub Khan","doi":"10.1109/UEMCON51285.2020.9298118","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298118","url":null,"abstract":"RTL verification is still one the most challenging activities in digital system development as it is still the bottleneck in the time-to-market for an integrated circuit development cycle. Thus reducing verification time is one of the most important targets. In this paper, a tool is developed to generate automatic tests from SystemVerilog assertions or SystemVerilog Coverage. The proposed tool is tested using different memory modules starting from single port RAM through Multiple ports RAM, FIFO and the DDRx families. The performance, regarding the runtime, has been compared with the handcrafted test case generation process. Moreover, the performance has been compared with other automatic test generation tools. Results shows the effectiveness of the proposed design. The proposed tool excelled in terms of its run-time, complexity, and coverage percentage.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130163169","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298129
Abhishek Shivanna, Sujan Ray, Khaldoon Alshouiliy, D. Agrawal
With the advancement of mobile and cloud technologies, there is a sharp increase in online transactions. Detecting fraudulent credit card transactions on a timely basis is a very critical and challenging problem in Financial Industry. Although online transactions are very convenient, they bring the risk of fraudulence on many aspects. Some of the key challenges in detecting fraudulence in online transactions include irregular behavioral patterns, skewed dataset i.e. high normal transaction to fraudulent transaction ratio, limited availability of data and dynamically changing environment. Every year people lose millions of dollars due to credit card fraud. There is a lack of quality research in this domain. We have used a dataset comprising of European cardholders which has 284,807 transactions to model our system. In this paper, we will design and develop credit card fraudulence detection system by training and testing two ML algorithms: Decision Forest (DF) and Decision Jungle (DJ) classifiers. Our results successfully demonstrate that DJ classifier delivers higher performance compared to DF classifier.
{"title":"Detection of Fraudulence in Credit Card Transactions using Machine Learning on Azure ML","authors":"Abhishek Shivanna, Sujan Ray, Khaldoon Alshouiliy, D. Agrawal","doi":"10.1109/UEMCON51285.2020.9298129","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298129","url":null,"abstract":"With the advancement of mobile and cloud technologies, there is a sharp increase in online transactions. Detecting fraudulent credit card transactions on a timely basis is a very critical and challenging problem in Financial Industry. Although online transactions are very convenient, they bring the risk of fraudulence on many aspects. Some of the key challenges in detecting fraudulence in online transactions include irregular behavioral patterns, skewed dataset i.e. high normal transaction to fraudulent transaction ratio, limited availability of data and dynamically changing environment. Every year people lose millions of dollars due to credit card fraud. There is a lack of quality research in this domain. We have used a dataset comprising of European cardholders which has 284,807 transactions to model our system. In this paper, we will design and develop credit card fraudulence detection system by training and testing two ML algorithms: Decision Forest (DF) and Decision Jungle (DJ) classifiers. Our results successfully demonstrate that DJ classifier delivers higher performance compared to DF classifier.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125306562","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298060
Kishor Datta Gupta, D. Dasgupta, Z. Akhtar
Modern deep learning models for the computer vision domain are vulnerable against adversarial attacks. Image prepossessing technique based defense against malicious input is currently considered obsolete as this defense is not effective against all types of attacks. The advanced adaptive attack can easily defeat pre-processing based defenses. In this paper, we proposed a framework that will generate a set of image processing sequences (several image processing techniques in a series). We randomly select a set of Image processing technique sequences (IPTS) dynamically to answer the obscurity question in testing time. This paper outlines methodology utilizing varied datasets examined with various adversarial data manipulations. For specific attack types and dataset, it produces unique IPTS. The outcome of our empirical experiments shows that the method can efficiently employ as processing for any machine learning models. The research also showed that our process works against adaptive attacks as we are using a non-deterministic set of IPTS for each adversarial input.
{"title":"Adversarial Input Detection Using Image Processing Techniques (IPT)","authors":"Kishor Datta Gupta, D. Dasgupta, Z. Akhtar","doi":"10.1109/UEMCON51285.2020.9298060","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298060","url":null,"abstract":"Modern deep learning models for the computer vision domain are vulnerable against adversarial attacks. Image prepossessing technique based defense against malicious input is currently considered obsolete as this defense is not effective against all types of attacks. The advanced adaptive attack can easily defeat pre-processing based defenses. In this paper, we proposed a framework that will generate a set of image processing sequences (several image processing techniques in a series). We randomly select a set of Image processing technique sequences (IPTS) dynamically to answer the obscurity question in testing time. This paper outlines methodology utilizing varied datasets examined with various adversarial data manipulations. For specific attack types and dataset, it produces unique IPTS. The outcome of our empirical experiments shows that the method can efficiently employ as processing for any machine learning models. The research also showed that our process works against adaptive attacks as we are using a non-deterministic set of IPTS for each adversarial input.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121983577","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298087
Ryoko Nino, T. Nishio, T. Murase
This paper demonstrates the feasibility of IEEE 802.11ad-based in-vehicle communication for a wireless harness. IEEE 802.11ad millimeter-wave (mmWave) communication enables high-speed wireless transmission, and its short communication range prevents harmful interference from other vehicles. However, in an in-vehicle environment, the received power of IEEE 802.11ad-based mmWave communications can be largely and easily attenuated by obstacles such as humans and the vehicle interior. Moreover, mmWave signals from adjacent vehicles can penetrate through vehicle windows and cause harmful interference. In this paper, we report the experimental results of in-vehicle communications using an actual vehicle and IEEE 802.11ad devices in an anechoic chamber. The experimental results demonstrate that IEEE 802.11ad-based in-vehicle communication can achieve a throughput of several hundred megabits per second, which is almost equivalent to that in achieved free space; this throughput can even be achieved when there are multiple obstacles in a vehicle and when adjacent vehicles (i.e., interferers) are in close proximity.
{"title":"IEEE 802.11ad Communication Quality Measurement in In-vehicle Wireless Communication with Real Machines","authors":"Ryoko Nino, T. Nishio, T. Murase","doi":"10.1109/UEMCON51285.2020.9298087","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298087","url":null,"abstract":"This paper demonstrates the feasibility of IEEE 802.11ad-based in-vehicle communication for a wireless harness. IEEE 802.11ad millimeter-wave (mmWave) communication enables high-speed wireless transmission, and its short communication range prevents harmful interference from other vehicles. However, in an in-vehicle environment, the received power of IEEE 802.11ad-based mmWave communications can be largely and easily attenuated by obstacles such as humans and the vehicle interior. Moreover, mmWave signals from adjacent vehicles can penetrate through vehicle windows and cause harmful interference. In this paper, we report the experimental results of in-vehicle communications using an actual vehicle and IEEE 802.11ad devices in an anechoic chamber. The experimental results demonstrate that IEEE 802.11ad-based in-vehicle communication can achieve a throughput of several hundred megabits per second, which is almost equivalent to that in achieved free space; this throughput can even be achieved when there are multiple obstacles in a vehicle and when adjacent vehicles (i.e., interferers) are in close proximity.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126319477","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298042
A. Lagunov, S. Zabolotniy
At present, the attention of many countries is directed to the Arctic. Many politicians argue about the boundaries in the circumpolar region. The problem is that the Arctic is very rich in hydrocarbons. Since there is practically no land in this region, it is necessary to produce hydrocarbons on the Arctic Ocean's sea shelf. For this, offshore platforms are used. It is complicated for a person to work on an offshore platform since the Arctic has shallow temperatures with extreme winds. At the same time, natural gas may appear inside the platform. Natural gas at specific concentrations mixed with oxygen can be explosive, increasing the risk of a person being on the platform. Therefore, to monitor the concentration of natural gas on the platform, special sensors register the gas level and give a control signal of danger. These sensors are not installed at all points. For those places where sensors are not installed, we have developed a particular mobile device that can determine the natural gas concentration and transmit it to the control device.
{"title":"Mobile Natural Gas Concentration Intelligence Device Study for the Arctic","authors":"A. Lagunov, S. Zabolotniy","doi":"10.1109/UEMCON51285.2020.9298042","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298042","url":null,"abstract":"At present, the attention of many countries is directed to the Arctic. Many politicians argue about the boundaries in the circumpolar region. The problem is that the Arctic is very rich in hydrocarbons. Since there is practically no land in this region, it is necessary to produce hydrocarbons on the Arctic Ocean's sea shelf. For this, offshore platforms are used. It is complicated for a person to work on an offshore platform since the Arctic has shallow temperatures with extreme winds. At the same time, natural gas may appear inside the platform. Natural gas at specific concentrations mixed with oxygen can be explosive, increasing the risk of a person being on the platform. Therefore, to monitor the concentration of natural gas on the platform, special sensors register the gas level and give a control signal of danger. These sensors are not installed at all points. For those places where sensors are not installed, we have developed a particular mobile device that can determine the natural gas concentration and transmit it to the control device.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"134 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128210333","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298137
A. Tolio, Davide Boem, Thomas Marchioro, L. Badia
LoRa networks have been gaining ground as a solution for Internet of Things because of their potential ability to handle massive number of devices. One of the most challenging problems of such networks is the need to set the Spreading Factors (SF) used by the terminals as close to a uniform distribution as possible, to guarantee reliable transmission of packets. This can be tackled through stochastic allocations based on centralized strategies, and more recently some contributions proposed fully distributed approaches based on game theory. However, these studies still consider games of complete information, where users have full knowledge on each other payoffs. In reality, it would be more appropriate to extend these approaches to Bayesian games, as we propose to do here. More precisely, we extend the game theoretic formulation to a semi-supervised allocation, where the distributed character of the allocation is retained as the nodes still act independently in choosing their SF, based on what they think it is their best preferred choice. We also utilize the central gateway as a coordinator regulating these proposals and the interaction of the nodes with the coordinator is framed as a Bayesian entry game, where nodes exploit a prior to decide whether to join the proposed allocation or not. Under this framework, nodes reach a satisfactory compromise between the assignment they receive from the network and their desired rate.
{"title":"A Bayesian Game Framework for a Semi-Supervised Allocation of the Spreading Factors in LoRa Networks","authors":"A. Tolio, Davide Boem, Thomas Marchioro, L. Badia","doi":"10.1109/UEMCON51285.2020.9298137","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298137","url":null,"abstract":"LoRa networks have been gaining ground as a solution for Internet of Things because of their potential ability to handle massive number of devices. One of the most challenging problems of such networks is the need to set the Spreading Factors (SF) used by the terminals as close to a uniform distribution as possible, to guarantee reliable transmission of packets. This can be tackled through stochastic allocations based on centralized strategies, and more recently some contributions proposed fully distributed approaches based on game theory. However, these studies still consider games of complete information, where users have full knowledge on each other payoffs. In reality, it would be more appropriate to extend these approaches to Bayesian games, as we propose to do here. More precisely, we extend the game theoretic formulation to a semi-supervised allocation, where the distributed character of the allocation is retained as the nodes still act independently in choosing their SF, based on what they think it is their best preferred choice. We also utilize the central gateway as a coordinator regulating these proposals and the interaction of the nodes with the coordinator is framed as a Bayesian entry game, where nodes exploit a prior to decide whether to join the proposed allocation or not. Under this framework, nodes reach a satisfactory compromise between the assignment they receive from the network and their desired rate.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"2006 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127335788","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298124
Jahnavi Kachhia, Rashika Natharani, K. George
The purpose of this paper is to combine Deep Learning with Brain-Computer Interface (BCI) for 3D Printing without human interference. This design will eliminate the intermediate steps and enable people to generate 3D prints faster. To collect the data, subjects are asked to wear g.Nautilus headsets and perform a mental imagery task. These collected brain waves are preprocessed using MATLAB and then are used to train different Neural Network architectures. The Neural Network model recognizes patterns in these brain waves to predict the shape imagined by the user. In this paper, we introduce CNN-LSTM that servers the purpose of classifying objects accurately. Once the shape is identified, the CAD file is generated in STL format using the predefined size. Lastly, this STL file is converted into G-code and serially transferred to the 3D Printer.
{"title":"Deep Learning Enhanced BCI Technology for 3D Printing","authors":"Jahnavi Kachhia, Rashika Natharani, K. George","doi":"10.1109/UEMCON51285.2020.9298124","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298124","url":null,"abstract":"The purpose of this paper is to combine Deep Learning with Brain-Computer Interface (BCI) for 3D Printing without human interference. This design will eliminate the intermediate steps and enable people to generate 3D prints faster. To collect the data, subjects are asked to wear g.Nautilus headsets and perform a mental imagery task. These collected brain waves are preprocessed using MATLAB and then are used to train different Neural Network architectures. The Neural Network model recognizes patterns in these brain waves to predict the shape imagined by the user. In this paper, we introduce CNN-LSTM that servers the purpose of classifying objects accurately. Once the shape is identified, the CAD file is generated in STL format using the predefined size. Lastly, this STL file is converted into G-code and serially transferred to the 3D Printer.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132098803","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298138
Y. Shah, S. Sengupta
Internet of Things (IoT) devices have gained popularity in recent years. With the increased usage of IoT devices, users have become more prone to Cyber-attacks. Threats against IoT devices must be analyzed thoroughly to develop protection mechanisms against them. An attacker’s purpose behind launching an attack is to find a weak link within a network and once discovered, the devices connected to the network become the primary target for the attackers. Industrial Internet of Things (IIoT) emerged due to the popularity of IoT devices and they are used to interconnect machines, sensors, and actuators at large manufacturing plants. By incorporating IIoT at their facilities companies have benefited by reducing operational costs and increasing productivity. However, as IIoT relies on utilizing the Internet to operate it is vulnerable to Cyber-attacks if security is not taken into consideration. After seeing the advantages of IIoT, a new version of smart industries has been introduced called Industry 4.0. Industry 4.0 combines cloud and fog computing, cyber-physical systems (CPS), and data analytics to automate the manufacturing process. This paper surveys the different classifications of attacks that an attacker can launch against these devices and mentions methods of mitigating such attacks1.
{"title":"A survey on Classification of Cyber-attacks on IoT and IIoT devices","authors":"Y. Shah, S. Sengupta","doi":"10.1109/UEMCON51285.2020.9298138","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298138","url":null,"abstract":"Internet of Things (IoT) devices have gained popularity in recent years. With the increased usage of IoT devices, users have become more prone to Cyber-attacks. Threats against IoT devices must be analyzed thoroughly to develop protection mechanisms against them. An attacker’s purpose behind launching an attack is to find a weak link within a network and once discovered, the devices connected to the network become the primary target for the attackers. Industrial Internet of Things (IIoT) emerged due to the popularity of IoT devices and they are used to interconnect machines, sensors, and actuators at large manufacturing plants. By incorporating IIoT at their facilities companies have benefited by reducing operational costs and increasing productivity. However, as IIoT relies on utilizing the Internet to operate it is vulnerable to Cyber-attacks if security is not taken into consideration. After seeing the advantages of IIoT, a new version of smart industries has been introduced called Industry 4.0. Industry 4.0 combines cloud and fog computing, cyber-physical systems (CPS), and data analytics to automate the manufacturing process. This paper surveys the different classifications of attacks that an attacker can launch against these devices and mentions methods of mitigating such attacks1.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125137835","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298167
Michael Guarino, Pablo Rivas, C. DeCusatis
On the frontier of cybersecurity are a class of emergent security threats that learn to find vulnerabilities in machine learning systems. A supervised machine learning classifier learns a mapping from x to y where x is the input features and y is a vector of associated labels. Neural Networks are state of the art performers on most vision, audio, and natural language processing tasks. Neural Networks have been shown to be vulnerable to adversarial perturbations of the input, which cause them to misclassify with high confidence. Adversarial perturbations are small but targeted modifications to the input often undetectable by the human eye. Adversarial perturbations pose risk to applications that rely on machine learning models. Neural Networks have been shown to be able to classify distributed denial of service (DDoS) attacks by learning a dataset of attack characteristics visualized using three-axis hive plots. In this work we present a novel application of a classifier trained to classify DDoS attacks that is robust to some of the most common, known, classes of gradient-based and gradient-free adversarial attacks.
{"title":"Towards Adversarially Robust DDoS-Attack Classification","authors":"Michael Guarino, Pablo Rivas, C. DeCusatis","doi":"10.1109/UEMCON51285.2020.9298167","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298167","url":null,"abstract":"On the frontier of cybersecurity are a class of emergent security threats that learn to find vulnerabilities in machine learning systems. A supervised machine learning classifier learns a mapping from x to y where x is the input features and y is a vector of associated labels. Neural Networks are state of the art performers on most vision, audio, and natural language processing tasks. Neural Networks have been shown to be vulnerable to adversarial perturbations of the input, which cause them to misclassify with high confidence. Adversarial perturbations are small but targeted modifications to the input often undetectable by the human eye. Adversarial perturbations pose risk to applications that rely on machine learning models. Neural Networks have been shown to be able to classify distributed denial of service (DDoS) attacks by learning a dataset of attack characteristics visualized using three-axis hive plots. In this work we present a novel application of a classifier trained to classify DDoS attacks that is robust to some of the most common, known, classes of gradient-based and gradient-free adversarial attacks.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133710813","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 : 2020-10-28DOI: 10.1109/UEMCON51285.2020.9298074
Pangao Du, Xianghong Lin, Xiaomei Pi, Xiangwen Wang
Deep recurrent spiking neural networks (DRSNNs) are stacked with the recurrent spiking neural machine (RSNM) modules. However, because of their intricately discontinuous and complex recurrent structures, it is difficult to pre-training the synaptic weights of RSNMs by simple and effective learning method in deep recurrent network. This paper proposed a new unsupervised multi-spike learning rule and the RSNM is trained by this rule, the complex spatiotemporal pattern of spike trains are learned. The spike signal will complete the two processes of forward propagation and reverse reconstruction, and then adjust the synaptic weight according to the error. This algorithm is successfully applied to spike trains, the learning rate and neuron number in the RSNMs are analyzed. In addition, the layer-wise pre-training method of DRSNN is presented, and the reconstruction error shows the algorithm has a better learning effect.
{"title":"An Unsupervised Learning Algorithm for Deep Recurrent Spiking Neural Networks","authors":"Pangao Du, Xianghong Lin, Xiaomei Pi, Xiangwen Wang","doi":"10.1109/UEMCON51285.2020.9298074","DOIUrl":"https://doi.org/10.1109/UEMCON51285.2020.9298074","url":null,"abstract":"Deep recurrent spiking neural networks (DRSNNs) are stacked with the recurrent spiking neural machine (RSNM) modules. However, because of their intricately discontinuous and complex recurrent structures, it is difficult to pre-training the synaptic weights of RSNMs by simple and effective learning method in deep recurrent network. This paper proposed a new unsupervised multi-spike learning rule and the RSNM is trained by this rule, the complex spatiotemporal pattern of spike trains are learned. The spike signal will complete the two processes of forward propagation and reverse reconstruction, and then adjust the synaptic weight according to the error. This algorithm is successfully applied to spike trains, the learning rate and neuron number in the RSNMs are analyzed. In addition, the layer-wise pre-training method of DRSNN is presented, and the reconstruction error shows the algorithm has a better learning effect.","PeriodicalId":433609,"journal":{"name":"2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"154 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133923668","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：