machine learning

Open Access Article

1 - Detecting Human Activities Based on Motion Sensors in IOT Using Deep Learning
Abbas Mirzaei fatemeh faraji

20.1001.1.16823745.1400.19.4.10.3

Control of areas and locations and motion sensors in the Internet of Things requires continuous control to detect human activities in different situations, which is an important challenge, including manpower and human error. Permanent human control of IoT motion sensors More

Control of areas and locations and motion sensors in the Internet of Things requires continuous control to detect human activities in different situations, which is an important challenge, including manpower and human error. Permanent human control of IoT motion sensors also seems impossible. The IoT is more than just a simple connection between devices and systems. IoT information sensors and systems help companies get a better view of system performance. This study presents a method based on deep learning and a 30-layer DNN neural network for detecting human activity on the Fordham University Activity Diagnostic Data Set. The data set contains more than 1 million lines in six classes to detect IoT activity. The proposed model had almost 90% and an error rate of 0.22 in the evaluation criteria, which indicates the good performance of deep learning in activity recognition. Manuscript profile

Open Access Article

2 - Deep Extreme Learning Machine: A Combined Incremental Learning Approach for Data Stream Classification
Javad Hamidzadeh Mona Moradi

20.1001.1.16823745.1401.20.1.5.9

Streaming data refers to data that is continuously generated in the form of fast streams with high volumes. This kind of data often runs into evolving environments where a change may affect the data distribution. Because of a wide range of real-world applications of dat More

Streaming data refers to data that is continuously generated in the form of fast streams with high volumes. This kind of data often runs into evolving environments where a change may affect the data distribution. Because of a wide range of real-world applications of data streams, performance improvement of streaming analytics has become a hot topic for researchers. The proposed method integrates online ensemble learning into extreme machine learning to improve the data stream classification performance. The proposed incremental method does not need to access the samples of previous blocks. Also, regarding the AdaBoost approach, it can react to concept drift by the component weighting mechanism and component update mechanism. The proposed method can adapt to the changes, and its performance is leveraged to retain high-accurate classifiers. The experiments have been done on benchmark datasets. The proposed method can achieve 0.90% average specificity, 0.69% average sensitivity, and 0.87% average accuracy, indicating its superiority compared to two competing methods. Manuscript profile

Open Access Article

3 - A Content-Based Image Retrieval System Using Semi-Supervised Learning and Frequent Patterns Mining
Maral Kolahkaj

20.1001.1.16823745.1401.20.3.7.5

Content-based image retrieval, which is also known as query based on image content, is one of the sub-branches of machine vision, which is used to organize and recognize the content of digital images using visual features. This technology automatically searches the imag More

Content-based image retrieval, which is also known as query based on image content, is one of the sub-branches of machine vision, which is used to organize and recognize the content of digital images using visual features. This technology automatically searches the images similar to the query image from huge image database and it provides the most similar images to the users by directly extracting visual features from image data; not keywords and textual annotations. Therefore, in this paper, a method is proposed that utilizes wavelet transformation and combining features with color histogram to reduce the semantic gap between low-level visual features and high-level meanings of images. In this regard, the final output will be presented using the feature extraction method from the input images. In the next step, when the query images are given to the system by the target user, the most similar images are retrieved by using semi-supervised learning that results from the combination of clustering and classification based on frequent patterns mining. The experimental results show that the proposed system has provided the highest level of effectiveness compared to other methods. Manuscript profile

Open Access Article

4 - SQ-PUF: A Resistant PUF-Based Authentication Protocol against Machine-Learning Attack
Abolfazl Sajadi Bijan Alizadeh

Physically unclonable functions (PUFs) provide hardware to generate a unique challenge-response pattern for authentication and encryption purposes. An essential feature of these circuits is their unpredictability, meaning that an adversary cannot sufficiently predict fu More

Physically unclonable functions (PUFs) provide hardware to generate a unique challenge-response pattern for authentication and encryption purposes. An essential feature of these circuits is their unpredictability, meaning that an adversary cannot sufficiently predict future responses from previous observations. However, machine learning algorithms have been demonstrated to be a severe threat to PUFs since they are capable of accurately modeling their behavior. In this work, we analyze PUF security threats and propose a PUF-based authentication mechanism called SQ-PUF, which can provide good resistance to machine learning attacks. In order to make it harder to simulate or predict, we obfuscated the correlation between challenge-response pairs. Experimental results show that, unlike existing PUFs, even with a large data set, the SQ-PUF model cannot be successfully attacked with a maximum prediction accuracy of 53%, indicating that this model is unpredictable. In addition, the uniformity in this model remains almost the same as the ideal value in A-PUF. Manuscript profile

Open Access Article

5 - Machine Learning-Based Security Resource Allocation for Defending against Attacks in the Internet of Things
Nasim Navaei Vesal Hakami

Nowadays, the Internet of Things (IoT) has become the focus of security attacks due to the limitation of processing resources, heterogeneity, energy limitation in objects, and the lack of a single standard for implementing security mechanisms. In this article, a solutio More

Nowadays, the Internet of Things (IoT) has become the focus of security attacks due to the limitation of processing resources, heterogeneity, energy limitation in objects, and the lack of a single standard for implementing security mechanisms. In this article, a solution will be presented for the problem of security resources allocating to deal with attacks in the Internet of Things. Security Resource Allocation (SRA) problem in the IoT networks refers to the placement of the security resources in the IoT infrastructure. To solve this problem, it is mandatory to consider the dynamic nature of the communication environments and the uncertainty of the attackers' actions. In the traditional approaches for solving the SRA, the attacker works over based on his assumptions about the system conditions. Meanwhile, the defender collects the system's information with prior knowledge of the attacker's behavior and the targeted nodes. Unlike the mentioned traditional approaches, this research has adopted a realistic approach for the Dynamic Security Resources Allocation in the IoT to battle attackers with unknown behavior. In the stated problem, since there is a need to decide on deploying several security resources during the learning periods, the state space of the strategies is expressed in the combinatorial form. Also, the SRAIoT problem is defined as a combinatorial-adversarial multi-armed bandit problem. Since switching in the security resources has a high cost, in real scenarios, this cost is included in the utility function of the problem. Thus, the proposed framework considers the switching cost and the earned reward. The simulation results show a faster convergence of the weak regret criterion of the proposed algorithms than the basic combinatorial algorithm. In addition, in order to simulate the IoT network in a realistic context, the attack scenario has been simulated using the Cooja simulator. Manuscript profile

Open Access Article

6 - Design and Collection of Speech Data as the First Step of Localization the Intelligent Diagnosis of Autism in Iranian Children
Maryam Alizadeh Shima tabibian

Autism Spectrum Disorder is a type of disorder in which, the patients suffer from a developmental disorder that manifests itself by symptoms such as inability to social communication. Thus, the most apparent sign of autism is a speech disorder. The first part of this pa More

Autism Spectrum Disorder is a type of disorder in which, the patients suffer from a developmental disorder that manifests itself by symptoms such as inability to social communication. Thus, the most apparent sign of autism is a speech disorder. The first part of this paper reviews research studies conducted to automatically diagnose autism based on speech processing methods. According to our review, the main speech processing approaches for diagnosing autism can be divided into two groups. The first group detects individuals with autism by processing their answers or feelings in response to questions or stories. The second group distinguishes people with autism from healthy people because of the accuracy of recognizing their spoken utterances based on automatic speech recognition systems. Despite much research being conducted outside Iran, few studies have been conducted in Iran. The most important reason for this is the lack of rich data that meet the needs of autism diagnosis based on the speech processing of suspected people. In the second part of the paper, we discuss the process of designing, collecting, and evaluating a speaker-independent dataset for autism diagnosis in Iranian children as the first step in the localization of the mentioned field. Manuscript profile

Open Access Article

7 - Combination of Instance Selection and Data Augmentation Techniques for Imbalanced Data Classification
Parastoo Mohaghegh Samira Noferesti Mehri Rajaei

Mohaghegh, S. Noferesti*, and M. Rajaei Abstract: In the era of big data, automatic data analysis techniques such as data mining have been widely used for decision-making and have become very effective. Among data mining techniques, classification is a common method fo More

Mohaghegh, S. Noferesti*, and M. Rajaei Abstract: In the era of big data, automatic data analysis techniques such as data mining have been widely used for decision-making and have become very effective. Among data mining techniques, classification is a common method for decision making and prediction. Classification algorithms usually work well on balanced datasets. However, one of the challenges of the classification algorithms is how to correctly predicting the label of new samples based on learning on imbalanced datasets. In this type of dataset, the heterogeneous distribution of the data in different classes causes examples of the minority class to be ignored in the learning process, while this class is more important in some prediction problems. To deal with this issue, in this paper, an efficient method for balancing the imbalanced dataset is presented, which improves the accuracy of the machine learning algorithms to correct prediction of the class label of new samples. According to the evaluations, the proposed method has a better performance compared to other methods based on two common criteria in evaluating the classification of imbalanced datasets, namely "Balanced Accuracy" and "Specificity". Manuscript profile

Current Issue

Published Issues

Menu

Browse

List of Articles machine learning