Pub Date : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973067
Swapnil Bhosale, I. Sheikh, Sri Harsha Dumpala, S. Kopparapu
Spoken Language Understanding (SLU) without speech-to-text conversion is more promising in low resource scenarios. These could be applications where there is not enough labeled data to train reliable speech recognition and language understanding systems, or where running SLU on edge is preferred over cloud based services. In this paper, we present an approach for building SLU without speech-to-text conversion in low resource scenarios using a transfer learning approach. We show that the intermediate layer representations from a pre-trained model outperforms the typically used Mel filter bank features. Moreover, the representations extracted from a model pre-trained on one language perform well even for SLU tasks on a different language.
{"title":"Transfer Learning for Low Resource Spoken Language Understanding without Speech-to-Text","authors":"Swapnil Bhosale, I. Sheikh, Sri Harsha Dumpala, S. Kopparapu","doi":"10.1109/IBSSC47189.2019.8973067","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973067","url":null,"abstract":"Spoken Language Understanding (SLU) without speech-to-text conversion is more promising in low resource scenarios. These could be applications where there is not enough labeled data to train reliable speech recognition and language understanding systems, or where running SLU on edge is preferred over cloud based services. In this paper, we present an approach for building SLU without speech-to-text conversion in low resource scenarios using a transfer learning approach. We show that the intermediate layer representations from a pre-trained model outperforms the typically used Mel filter bank features. Moreover, the representations extracted from a model pre-trained on one language perform well even for SLU tasks on a different language.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121203135","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973064
Manan Mehta, Anmol Sajnani, Radhika Chapaneri
A cover song, by definition, is a rendition of a previously released song and mapping these cover songs to their original song is defined as ”Cover Song Identification.” In this paper, we propose multiple cover song identification methods using Convolutional Neural Network (CNN) models as well as transfer learning to extract features which can be trained on statistical models for binary classification. We develop two CNN models that are trained on a cross-similarity matrix which is generated from a pair of songs as input. Firstly we designed a simple CNN architecture that was trained on two labels 1. cover pair relationship; 2. non-cover pair relationship. Our second approach uses a CNN model known as the Inception Model. We train the model by generating cross-similarity matrices for both the labels and then converting them into images. At later stage, we use a ranking method that sorts the probabilities of the cover relation in descending order and the song with the highest probability is chosen as a match. Based on the evaluation, Inception model performs the best, scoring the accuracy of 93.4%.
{"title":"Cover Song Identification with Pairwise Cross-Similarity Matrix using Deep Learning","authors":"Manan Mehta, Anmol Sajnani, Radhika Chapaneri","doi":"10.1109/IBSSC47189.2019.8973064","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973064","url":null,"abstract":"A cover song, by definition, is a rendition of a previously released song and mapping these cover songs to their original song is defined as ”Cover Song Identification.” In this paper, we propose multiple cover song identification methods using Convolutional Neural Network (CNN) models as well as transfer learning to extract features which can be trained on statistical models for binary classification. We develop two CNN models that are trained on a cross-similarity matrix which is generated from a pair of songs as input. Firstly we designed a simple CNN architecture that was trained on two labels 1. cover pair relationship; 2. non-cover pair relationship. Our second approach uses a CNN model known as the Inception Model. We train the model by generating cross-similarity matrices for both the labels and then converting them into images. At later stage, we use a ranking method that sorts the probabilities of the cover relation in descending order and the song with the highest probability is chosen as a match. Based on the evaluation, Inception model performs the best, scoring the accuracy of 93.4%.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128479309","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8972992
K. Manasvi Bhat, P. Anchalia, S. Yashashree, R. Sanjeetha, A. Kanavalli
Epilepsy, a disorder that leads to abnormal activities in the brain is primarily caused by excessive neuronal activity. Patients diagnosed with epilepsy frequently suffer from seizures, the impact of which may vary from abnormal body movements to alterations in the levels of consciousness. An appropriate dosage of medication provided at the right time can help prevent an impending seizure. In this paper, real data obtained from Epilepsy Ecosystem is used for analysis. After preprocessing this data, several signal processing algorithms and mathematical computations are used for feature extraction. Two sets of features are identified viz. lasting features and transitory features. Several combinations of these features along with Machine Learning algorithms such as Extra Trees Classifier and XGBoost are used to train generalized models as well as a patient-specific models, both of which are immune to noise. It is observed that the XGBoost based generalized model which is trained using lasting features gives a relatively better accuracy of 90.41%.
癫痫是一种导致大脑异常活动的疾病，主要是由过度的神经元活动引起的。被诊断为癫痫的患者经常遭受癫痫发作，其影响可能从异常的身体运动到意识水平的改变。在适当的时间给予适当剂量的药物可以帮助预防即将发生的癫痫发作。本文采用癫痫生态系统的真实数据进行分析。在对该数据进行预处理后，采用多种信号处理算法和数学计算进行特征提取。确定了两组特征，即持久特征和短暂特征。这些特征的几种组合以及机器学习算法(如Extra Trees Classifier和XGBoost)被用于训练广义模型和特定患者模型，这两种模型都不受噪声的影响。观察到，使用持久特征训练的基于XGBoost的广义模型的准确率相对较高，为90.41%。
{"title":"Detection and Prediction of the Preictal State of an Epileptic Seizure using Machine Learning Techniques on EEG Data","authors":"K. Manasvi Bhat, P. Anchalia, S. Yashashree, R. Sanjeetha, A. Kanavalli","doi":"10.1109/IBSSC47189.2019.8972992","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8972992","url":null,"abstract":"Epilepsy, a disorder that leads to abnormal activities in the brain is primarily caused by excessive neuronal activity. Patients diagnosed with epilepsy frequently suffer from seizures, the impact of which may vary from abnormal body movements to alterations in the levels of consciousness. An appropriate dosage of medication provided at the right time can help prevent an impending seizure. In this paper, real data obtained from Epilepsy Ecosystem is used for analysis. After preprocessing this data, several signal processing algorithms and mathematical computations are used for feature extraction. Two sets of features are identified viz. lasting features and transitory features. Several combinations of these features along with Machine Learning algorithms such as Extra Trees Classifier and XGBoost are used to train generalized models as well as a patient-specific models, both of which are immune to noise. It is observed that the XGBoost based generalized model which is trained using lasting features gives a relatively better accuracy of 90.41%.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125258480","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973033
Tanmay Chakraborty, Akash Kumar
Manually controlling an UAV can be a tough task with so many control parameters involved like altitude of the UAV, position of the UAV, velocity of the UAV. This work sheds light on a novel control architecture for easier and broader range control of an UAV using the GSM network with help of DTMF signals, Propeller Thrust equation, and Sensor Fusion with sensor reading compensation technique. We have fused a low cost accelerometer and gyroscope sensor with barometer sensor for calculating vertical velocity and altitude of the UAV from the sensor readings using an adaptive Kalman Filter. The whole system is integrated in the proposed flight controller architecture.
{"title":"Adaptive Sensor Fusion and Propeller Thrust Equation Based Digital Control System for UAV","authors":"Tanmay Chakraborty, Akash Kumar","doi":"10.1109/IBSSC47189.2019.8973033","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973033","url":null,"abstract":"Manually controlling an UAV can be a tough task with so many control parameters involved like altitude of the UAV, position of the UAV, velocity of the UAV. This work sheds light on a novel control architecture for easier and broader range control of an UAV using the GSM network with help of DTMF signals, Propeller Thrust equation, and Sensor Fusion with sensor reading compensation technique. We have fused a low cost accelerometer and gyroscope sensor with barometer sensor for calculating vertical velocity and altitude of the UAV from the sensor readings using an adaptive Kalman Filter. The whole system is integrated in the proposed flight controller architecture.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127858524","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973066
Divya K. Sawant, Anchal Jaiswal, Jyoti Singh, Payal Shah
In India, agriculture plays a predominant role in economy and employment. The common problem existing among the Indian farmers today is that they fail to choose the right crop based on their region specications and yield history. Hence they face a serious setback in productivity. Agricultural statistics and forecast is an important resource that the government has not explored commensurate to its impact. The paper proposes an intelligent portable system using data mining and analytics which assists farmers with various farming techniques, helps them decide most suitable crops as per current climate conditions, soil conditions and geographical characteristics of the specified region.The farmers do not have a single source which can cater to all their queries regarding seeds, fertilizers, market prices, storage facilities, government schemes,etc. To make this data analysis easily accessible to the farmers a chatbot is proposed which uses the Natural Language Processing technique. It helps to get responses of the farmer input queries regarding agricultural context in audio format, so as to make farmer interaction more user friendly. If the system fails to answer any specified query, the query is redirected to helpline centers. The system basically works as a virtual, handy assistant to assist farmers throughout the year helping them stay notified of any factor that would affect crop productivity and profit. The responses are generated based on various machine learning algorithms modelled around data set. Though the main audience under consideration are farmers any other user can also use the system to get advice regarding activities related to agriculture.
{"title":"AgriBot - An intelligent interactive interface to assist farmers in agricultural activities","authors":"Divya K. Sawant, Anchal Jaiswal, Jyoti Singh, Payal Shah","doi":"10.1109/IBSSC47189.2019.8973066","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973066","url":null,"abstract":"In India, agriculture plays a predominant role in economy and employment. The common problem existing among the Indian farmers today is that they fail to choose the right crop based on their region specications and yield history. Hence they face a serious setback in productivity. Agricultural statistics and forecast is an important resource that the government has not explored commensurate to its impact. The paper proposes an intelligent portable system using data mining and analytics which assists farmers with various farming techniques, helps them decide most suitable crops as per current climate conditions, soil conditions and geographical characteristics of the specified region.The farmers do not have a single source which can cater to all their queries regarding seeds, fertilizers, market prices, storage facilities, government schemes,etc. To make this data analysis easily accessible to the farmers a chatbot is proposed which uses the Natural Language Processing technique. It helps to get responses of the farmer input queries regarding agricultural context in audio format, so as to make farmer interaction more user friendly. If the system fails to answer any specified query, the query is redirected to helpline centers. The system basically works as a virtual, handy assistant to assist farmers throughout the year helping them stay notified of any factor that would affect crop productivity and profit. The responses are generated based on various machine learning algorithms modelled around data set. Though the main audience under consideration are farmers any other user can also use the system to get advice regarding activities related to agriculture.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134390264","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973004
Sumit Kumar Nagdeo, Judhistir Mahapatro
Sensor Networks are very much vulnerable and prone to faults and external attacks. Sensor networks used for Healthcare Monitoring are termed as Wireless Body Area Networks (WBAN), which is used for collecting various vital physiological parameters of patients from remote locations. However, WBAN sensors are prone to failures because of noise, hardware misplacement, patient‘s sweating. Sensed data from these sensors are sent from the Local Processing Unit to Medical Professionals. It would be very difficult for the Medical Professionals to diagnose correctly if the sensed data from these sensors are faulty or effected by the malicious third party. At times, even faulty data might lead to misdiagnosis or death of a patient. It motivated us to address this challenge by proposing a Machine Learning Paradigm to distinguish this anomalous data from the genuine sensed data. Firstly, we classify the health parameters as normal records or abnormal record. After the classification, we propose to apply regression technique for identifying the anomalous data and actual critical data. We use real patient‘s vital physiological parameters for validating the robustness and reliability of our proposed approach.
{"title":"Wireless Body Area Network Sensor Faults and Anomalous Data Detection and Classification using Machine Learning","authors":"Sumit Kumar Nagdeo, Judhistir Mahapatro","doi":"10.1109/IBSSC47189.2019.8973004","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973004","url":null,"abstract":"Sensor Networks are very much vulnerable and prone to faults and external attacks. Sensor networks used for Healthcare Monitoring are termed as Wireless Body Area Networks (WBAN), which is used for collecting various vital physiological parameters of patients from remote locations. However, WBAN sensors are prone to failures because of noise, hardware misplacement, patient‘s sweating. Sensed data from these sensors are sent from the Local Processing Unit to Medical Professionals. It would be very difficult for the Medical Professionals to diagnose correctly if the sensed data from these sensors are faulty or effected by the malicious third party. At times, even faulty data might lead to misdiagnosis or death of a patient. It motivated us to address this challenge by proposing a Machine Learning Paradigm to distinguish this anomalous data from the genuine sensed data. Firstly, we classify the health parameters as normal records or abnormal record. After the classification, we propose to apply regression technique for identifying the anomalous data and actual critical data. We use real patient‘s vital physiological parameters for validating the robustness and reliability of our proposed approach.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132131547","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973068
Rajarshi Mukhopadhyay, Soutrik Bandyopadhyay, A. Sutradhar, P. Chattopadhyay
Use of Reinforcement Learning (RL) in designing adaptive self-tuning PID controllers is a relatively new horizon of research with Q-learning and its variants being the predominant algorithms found in the literature. However, the possibility of using an interesting alternative algorithm i.e. Advantage Actor Critic (A2C) in the above context is relatively unexplored. In the present study, Deep Q Networks (DQN) and A2C approaches have been employed to design self-tuning PID controllers. Comparative performance analysis of both the controllers was undertaken in a simulation environment on a servo position control system, with various static and dynamic control objectives, keeping a conventional PID controller as a baseline. A2C based Adaptive PID Controller(A2CAPID) is more promising in trajectory tracking problems whereas DQN based Adaptive PID Controller(DQNAPID) is rather suitable for systems with relatively large plant parameter variations.
{"title":"Performance Analysis of Deep Q Networks and Advantage Actor Critic Algorithms in Designing Reinforcement Learning-based Self-tuning PID Controllers","authors":"Rajarshi Mukhopadhyay, Soutrik Bandyopadhyay, A. Sutradhar, P. Chattopadhyay","doi":"10.1109/IBSSC47189.2019.8973068","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973068","url":null,"abstract":"Use of Reinforcement Learning (RL) in designing adaptive self-tuning PID controllers is a relatively new horizon of research with Q-learning and its variants being the predominant algorithms found in the literature. However, the possibility of using an interesting alternative algorithm i.e. Advantage Actor Critic (A2C) in the above context is relatively unexplored. In the present study, Deep Q Networks (DQN) and A2C approaches have been employed to design self-tuning PID controllers. Comparative performance analysis of both the controllers was undertaken in a simulation environment on a servo position control system, with various static and dynamic control objectives, keeping a conventional PID controller as a baseline. A2C based Adaptive PID Controller(A2CAPID) is more promising in trajectory tracking problems whereas DQN based Adaptive PID Controller(DQNAPID) is rather suitable for systems with relatively large plant parameter variations.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"152 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127143368","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973080
Sachin Sen, C. Jayawardena
The information and networking technology have been revolutionized with the inception of recent evolution of Cyber-Physical Systems (CPS) and the Internet of Things (IoT). The next generation distributed computing systems i.e., CPS and IoT are highly interconnected and deeply embedded with the physical world. By capitalizing the advantages and opportunities of these technologies, Industrial-IoT has been fueling smart industrial processes. The execution and application of these processes generate huge amount of data, which leads the distributed computing systems to carefully consider information and data management reliably and securely; also facilitating the necessary automation as well as ensuring timely information exchange. But the current internet doesn’t guarantee the network performance and secure transportation; in addition the physical systems are becoming more insecure when interconnected to the cyber systems. These bottlenecks are leading to the necessity of improving performance and security in the cyber-physical communication. Considering those pervasive requirements, this paper has modeled network performance as well as system security with a view to improve these components which could heel the reliable cyber-communication challenges.
{"title":"Cybersecurity and Network Performance Modeling in Cyber-Physical Communication for BigData and Industrial IoT Technologies","authors":"Sachin Sen, C. Jayawardena","doi":"10.1109/IBSSC47189.2019.8973080","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973080","url":null,"abstract":"The information and networking technology have been revolutionized with the inception of recent evolution of Cyber-Physical Systems (CPS) and the Internet of Things (IoT). The next generation distributed computing systems i.e., CPS and IoT are highly interconnected and deeply embedded with the physical world. By capitalizing the advantages and opportunities of these technologies, Industrial-IoT has been fueling smart industrial processes. The execution and application of these processes generate huge amount of data, which leads the distributed computing systems to carefully consider information and data management reliably and securely; also facilitating the necessary automation as well as ensuring timely information exchange. But the current internet doesn’t guarantee the network performance and secure transportation; in addition the physical systems are becoming more insecure when interconnected to the cyber systems. These bottlenecks are leading to the necessity of improving performance and security in the cyber-physical communication. Considering those pervasive requirements, this paper has modeled network performance as well as system security with a view to improve these components which could heel the reliable cyber-communication challenges.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114728709","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}
Chatbots have effectively reduced human efforts by providing automated human-like solutions for various business and societal problems. This paper is an elaborate description of the design and implementation of a University Counselling Auto-Reply Bot, that is capable of providing answers to queries related to the field of Engineering at our University level. The appropriate NLP techniques are applied to our University Data, developed in the JSON format and the Feedforward neural model is used for training this dataset, the issue of overfitting was handled. The Chat Application is then deployed on Facebook Messenger, the response is visible to the User on the Facebook Messenger interface, to provide them with an effective interaction platform. The end-user testing was conducted in two phases; the probability scores of the correct responses were improved to 0.72 in the second phase from 0.46 in the first phase after devising additional training phrases and keywords to the dataset.
{"title":"Artificial Neural Network Based University Chatbot System","authors":"Namrata Bhartiya, Namrata Jangid, Sheetal Jannu, Purvika Shukla, Radhika Chapaneri","doi":"10.1109/IBSSC47189.2019.8973095","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973095","url":null,"abstract":"Chatbots have effectively reduced human efforts by providing automated human-like solutions for various business and societal problems. This paper is an elaborate description of the design and implementation of a University Counselling Auto-Reply Bot, that is capable of providing answers to queries related to the field of Engineering at our University level. The appropriate NLP techniques are applied to our University Data, developed in the JSON format and the Feedforward neural model is used for training this dataset, the issue of overfitting was handled. The Chat Application is then deployed on Facebook Messenger, the response is visible to the User on the Facebook Messenger interface, to provide them with an effective interaction platform. The end-user testing was conducted in two phases; the probability scores of the correct responses were improved to 0.72 in the second phase from 0.46 in the first phase after devising additional training phrases and keywords to the dataset.","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124768296","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 : 2019-07-01DOI: 10.1109/IBSSC47189.2019.8973099
Tanmay Toraskar, Ujwala M. Bhangale, Suchitra Patil, Neelkamal More
In this digital era, the number of Cybercrimes is increasing that has resulted in increased number of pending cybercrimes cases such as artifacts as a malware, hacking and cyber fraud or e-harassment. In order to deal with these cases, digital forensics must include the concrete law enforcement in the court of law. In digital forensics, it is challenging task to detect reliable evidence because of worldwide use and advancements in digital communication technologies.Common approaches such as file signature analysis and the data carving can be done using the forensics tools, however, digital evidence examiners are keen to find the relevant data which helps in finding the truth behind the case. To reduce the examination time in the data examination or analysis process, this paper explores the role of unsupervised pattern recognition to identify the notable artefact. The Self-Organising Map (SOM) is used to automatically cluster notable artefacts. In this work, four cases are presented to demonstrate the use of SOM in examining the digital data saved in a CSV format. Multiple SOMs are created including Extension Mismatch SOM that represents the intentional changes done on the default extension of the file in order to hide it from the forensic examiner. Other types of SOM are created for the EXIF Metadata (i.e. MAC attributes). USB Device Attached (Device Make, Device Model, Device ID, Date/Time, Source File, Tags).
{"title":"Efficient Computer Forensic Analysis Using Machine Learning Approaches","authors":"Tanmay Toraskar, Ujwala M. Bhangale, Suchitra Patil, Neelkamal More","doi":"10.1109/IBSSC47189.2019.8973099","DOIUrl":"https://doi.org/10.1109/IBSSC47189.2019.8973099","url":null,"abstract":"In this digital era, the number of Cybercrimes is increasing that has resulted in increased number of pending cybercrimes cases such as artifacts as a malware, hacking and cyber fraud or e-harassment. In order to deal with these cases, digital forensics must include the concrete law enforcement in the court of law. In digital forensics, it is challenging task to detect reliable evidence because of worldwide use and advancements in digital communication technologies.Common approaches such as file signature analysis and the data carving can be done using the forensics tools, however, digital evidence examiners are keen to find the relevant data which helps in finding the truth behind the case. To reduce the examination time in the data examination or analysis process, this paper explores the role of unsupervised pattern recognition to identify the notable artefact. The Self-Organising Map (SOM) is used to automatically cluster notable artefacts. In this work, four cases are presented to demonstrate the use of SOM in examining the digital data saved in a CSV format. Multiple SOMs are created including Extension Mismatch SOM that represents the intentional changes done on the default extension of the file in order to hide it from the forensic examiner. Other types of SOM are created for the EXIF Metadata (i.e. MAC attributes). USB Device Attached (Device Make, Device Model, Device ID, Date/Time, Source File, Tags).","PeriodicalId":148941,"journal":{"name":"2019 IEEE Bombay Section Signature Conference (IBSSC)","volume":"162 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120913608","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：