clustering

Vehicular ad hoc networks is one of the examples of intelligentization is the development of transportation infrastructure , which has led to the creation of new architectures based on cloud and fog computing. Mobilility and high speed of vehicles lead to instability in these types of networks and create obstacles in reliable information sharing. Clustering-based solutions are an optimal solution for network stability, achieving various features such as service quality and information dissemination. Among the issues that play an effective role in the stability of the cluster, we can mention how to build the cluster, how to choose the cluster head, and how to transfer data. In other words, one of most important challenges in these networks is how to cluster and select the cluster head. One of the main disadvantages of clustering algorithms is cluster instability and inappropriate cluster head selection. In this research, a two-level clustering method for case networks between vehicles is presented, whose lower level (the level close to the cars) is named as fog and its upper level (the level close to the infrastructure) is named as the cloud. Local data transmission and location management structures are performed at the fog level, and other operations related to data transfer to the infrastructure and sometimes distributed computing are performed through the cloud. The proposed method performs clustering with the criteria of speed, direction and location, and increases the stability of the cluster by selecting the appropriate cluster head. In addition, by providing a routing algorithm based on predicting the way cars move, the load is divided between the cluster head, the cloud, and the car. It can simultaneously improve the overhead and service quality of the inter-vehicle ad hoc network. Another disadvantage of the methods is the delay in sending information. The proposed method reduces the delay by making predictions and sending information correctly. The results of simulations show that compared to similar methods, the proposed method uses fewer steps to send information and improves the delay, delivery rate, and clustering overhead by 41, 18, and 29 percent respectively.

The Internet of Things (IoT) is a rapidly evolving technology that connects physical devices through networked systems. However, as IoT continues to expand, it poses various security challenges that require appropriate solutions to protect sensitive information and user privacy. This paper focuses on improving the speed of intrusion detection systems (IDS) as a critical solution for IoT security. In IDS, the large volume of data can slow down the learning process. In this paper, the DBSCAN clustering algorithm is modified by adding a minimum neighborhood parameter to reduce data samples in a targeted manner, aiming to enhance the speed of IDS and reduce learning time and costs. The parameters of the modified DBSCAN are tuned using a genetic algorithm. Experimental results on the Kaggle and NSL_KDD datasets demonstrate that the proposed model can maintain classification accuracy above 96% for the Kaggle dataset and above 92.51% for the NSL_KDD dataset, even with up to an 80% reduction in data volume. Additionally, computation time for the Kaggle dataset decreased from 458.09 ms to 47.21 ms, and for the NSL_KDD dataset from 995.2 ms to 223.60 ms. Thus, despite improvements in speed and reductions in time and cost, the model's optimal performance is maintained.

This research has been done with the aim of providing a routing method in Internet of Things networks based on fuzzy logic and clustering for the use of smart transportation. Today, transportation is one of the basic issues and major problems in cities and this issue bigger cities such as Tehran have become an important concern, hence transportation systems have been created and expanded. But rapid population growth, migration to cities, increasing car ownership and the use of personal vehicles are the challenges that urban transportation systems are facing. Meanwhile, passenger transportation systems can play a significant role in improving the city's traffic situation. Cluster-based routing protocols are one of the most important ways to reduce energy consumption in wireless sensor networks. In this thesis, a new clustering protocol using bee colony optimization algorithm is presented and a new method to reduce energy consumption and. Lifetime improvement in wireless sensor network routing includes three stages of clustering sensor nodes into a number of clusters using fuzzy clustering algorithm, selecting optimal cluster heads and data transmission. The results of implementing the proposed method and comparing its performance with other methods the former showed the proper performance of this method in this field

Nowadays, detecting unusual events in the network has been the subject of many researches. Network traffic is huge and very large, and this leads to high data size and increased noise, which makes it very difficult to extract meaningful information to detect abnormal events. Early detection of attacks improves the stability of a system. Each attack is a type of specific behavior; But some attacks may behave similarly and differ only in some features. This article presents a new way to detect malware and attacks in the cloud computing environment. In this method, data clustering separates the data from each other to provide better conditions for model construction by balancing the data in different classes. This research uses a combination of Adabost, Random Forest and Bosted Gradient Tree algorithms as ensemble learning to improve malware detection in cloud computing. In order to combine boosted learners and build a higher level model, the voting mechanism is used. In the proposed model, ensemble learning, using the strengths of various algorithms, creates a useful, high-performance system for detecting malware in cloud computing. By applying the proposed method on real data, it was observed that the accuracy of the proposed method is equal to 99.96%, its accuracy is equal to 99.97% and its recall is equal to 99.95% which compared to previous methods, it has a noticeable advantage, but its computational complexity has not changed much.

Selecting the appropriate cluster head nodes as well as determining the correct radius for the clusters are two key issues in ensuring the performance of cluster-based Wireless Sensor Networks (WSNs). In this paper, a routing and clustering algorithm for wireless sensor network is presented. The clustering algorithm presented in this research uses unequal clustering technique. This means that in the clustered structure of the network, the size of each cluster may differ from others. This structure reduces energy consumption in crowded areas by using clusters with smaller radius, and increase network throughput by using larger radius for clusters located in areas with low traffic. In the proposed method, the multi-objective cuckoo search algorithm is used to determine the optimal cluster nodes and also to determine the optimal radius for each cluster. After determining the clustered structure of the network, a routing algorithm based on game theory is used to determine the optimal paths for sending data to the base station. The performance of the proposed method in a simulated environment is evaluated and its efficiency is compared with previous algorithms. The simulation results show that by using the proposed method, in addition to reducing energy consumption, network traffic can be prevented and load distribution can be done more efficiently.

One of the most popular ways to detect malware is to find a match for malware file signature pattern in the malware signature database. The malware signature database is pre-extracted and is constantly updated. Checking the similarity of input data using the stored signatures causes storage problems and increases the calculation costs. In addition, the detection based on adapting the malware signature pattern fails when changing the malware code in polymorphic malware. In this paper, by combining the static analysis of executable file structure and the machine learning algorithms, an effective method for malware detection is presented. The data set for training and evaluation of the proposed method includes 36,567 samples of malware and 17295 benign files, and the malware is clustered in 7 families. The results show that the presented method is able to detect and cluster malware from benign files with an accuracy of more than 99% and a false positive rate less than 0.4%. The proposed method has very low processing overheads compared to similar methods and the average scanning time of executable files is 0.244 second.

The main purpose of wireless sensor networks (WSNs) is to monitor, record and announce specific conditions from different locations and different applications to the well node or end user. Wireless sensor networks have many applications such as patient status monitoring, target tracking, forest and rangeland monitoring, battlefield, and so on. In these networks, energy is one of the inherent limitations. Because the energy consumed is supplied by a battery, which has a limited lifespan. Clustering is one of the best ways to save energy due to data aggregation, and selecting the right clusters increases the lifespan of wireless sensor networks. Since clustering is one of the NP-hard problems, metaheuristic algorithms are suitable for this problem. In this paper, an energy-aware and cluster-based routing method for WSNs with a combination of multilayer perceptron (MLP)neural network algorithm and simulated annealing (SA) is presented. In the proposed method, the simulated annealing metaheuristic algorithm is simulated to determine the cluster head (CH) and multilayer perceptron neural networks are used to determine the members of each cluster. After the clustering process, data is sent from the source node to the well by creating appropriate routing tables among the headers. The simulation results of the proposed method show that this method improves the parameters of energy consumption, package delivery rate and throughput.

In today's critical situation, wireless sensor networks have attracted much attention as one of the newest communication technologies for use in passive defense applications. In this regard, the limited energy  resources of these networks and consequently their limited network lifetime, is a serious problem in  applying them for such critical applications. Recently with the advent of energy harvesting methods, it has become possible to overcome this problem. In this paper, a new cluster-based multipath routing protocol called R-DEARER is proposed for use in energy harvesting wireless sensor networks. In this protocol,  inter-cluster routing is used to make possible the multi-hop communication between cluster heads and the sink. For this purpose, the cluster-heads send their collected data packets to relay cluster-heads that are near the sink and then the relay cluster-heads send packets to the sink. By using this method, the energy consumption of the cluster-head nodes will be reduced; therefore, the network lifetime and energy efficiency will be increased significantly. The performance of the proposed protocol is examined by analyzing and evaluating the numerical results. The results show that in terms of energy efficiency, packet delivery rate, and network outage probability, the proposed R-DEARER protocol performs better than the same method in which the communication from cluster-heads to the sink is single hop.

Database outsourcing in cloud computing is one of the main solutions to maintain and access a database easily without the need for large infrastructure investment. Although data owners usually trust service  providers and infrastructure providers in terms of maintainability and accessibility, but they are concerned about the privacy and confidentiality of information, and therefore prefer to keep data encrypted on cloud servers. Encrypted data is not searchable, and a solution needs to be provided by the server to search for that data. One solution is to use keyword indexing as metadata alongside the encrypted database. There are several key challenges to using these solutions: high volume of indexes, user error exposure and search speed. In this study, fuzzy keyword sets are used instead of fixed words when searching for users, and in order to reduce storage space using the keyword clustering method, appropriate fuzzy keyword sets are  selected and metadata with less production volume is used and encrypted. Using hierarchical clustering methods with specific metrics, the same keywords are placed in a cluster, and to find the desired keyword, it is not necessary to search all metadata and thus the search time is reduced. Practical results and evaluations show that the proposed method is practical, safe and efficient.

Today, with the spread of social networks, the opposition's efforts to chill out people from government (known as “soft war”) are increased. Therefore, dealing with this type of networks is important for military and security organizations. Graph clustering is one of the first attempts toward analyzing social networks which can appropriately be modeled by a content graph. In contrast, most of the existing graph clustering methods independently focused on one of the content or structural aspects of a graph. The aim of this paper (implemented as CS-Cluster algorithm) is to achieve well connected clusters while their nodes benefits from homogeneous attribute values (content). In the second step of our research, after an intensive search, no measure has found which could simultaneously consider content and structural features of clustering algorithms. So to be able to appropriately evaluate our algorithm, a new content-structural measure (so-called “CS-Measure”) is proposed. Our experimentation shows that the proposed clustering algorithm outperforms two other well-known content-structural clustering algorithms, using the new content-structural, average similarity, and Error link measure as well as the previous content and structural measures, And it also performed relatively well in density measure.

Designing a suitable camouflage pattern for a building is one of the fundamental challenges in the field of passive defence. In this study, to design visible camouflage patterns, an automatic method based on high resolution satellite images processing is proposed. In the proposed method, the area around the target building on the image is first removed. Then, using the surrounding of the removed part, an artificial image is reconstructed by a texture extraction method. In the next step, the main colors of the image are extracted from reproduced image using K-means clustering algorithm, and then, the camouflage pattern is created by assigning the color of center of each cluster to all pixels of that cluster. The obtained camouflage pattern is evaluated on two different dataset in both visual assessment and computation the saliency value. Successful visual assessment and the 27 percent decrease in obtained saliency value using the second dataset indicate the high ability of the proposed method to camouflage buildings.

Wireless sensor networks afford the possibility to control remote monitoring of many environments (such as military ones) by hundreds of tiny sensor nodes. Restriction of energy consumption is a major challenge in these networks, which will affect the lifetime of the network. One of the key solutions to solving the challenge is data aggregation and avoiding of repeated data sending. The most popular communication protocols are clustering and tree based data aggregation. Clustering in data aggregation issue leads to energy balance, but energy consumption is high due to long distances between cluster heads and base station. In the tree structure, due to short distances between nodes, energy consumption is low but, the depth of the tree is usually high. In this paper, a hybrid analytical hierarchical process named CTDA is proposed in which energy consumption is reduced by clustering and minimum spanning tree for data aggregation in wireless sensor networks. The simulation of the proposed method illustrates reduction in energy consumption compared to two aformentioned protocols.

In parallel with advances in robots hardware, the motion planning science has become increasingly indispensable and complicated, and enables robots to enhance their capabilities in applications where the presence of humans is impractical or hazardous. The covering and minesweeping operations of outdoor areas are examples of such applications. This paper deals with the problem of covering and minesweeping of unknown environments by multiple mobile robots which have cooperative interactions. For searching the space, the new method of multi sensor-based random trees (MSRT) is developed, and for planning the robots’ paths toward the scattered items and their collection, a mathematical model similar to the multiple traveling salesman problem is proposed, which together with the k-means clustering, the A* search, and the visibility graph techniques plans the route of each robot. Comparisons with the counterpart mathematic model solved to optimality showed that the proposed method produces suboptimal solutions in much shorter computational times.

Audio Segmentation has got a widespread application in underwater signal processing specially observation of passing objects using their acoustic signals and also in audio annotation of a recorded meeting, which also used by intelligence services that is one of instances of passive defense.Audio Segmentation is the process of partitioning an audio stream into regions each of which corresponds to only one audio source or one speaker.There are various methods for speaker segmentation the most common methods of which are based on BIC criteria. These methods needs heavy statistical computations and are very time consuming.The main goal of this paper is to propose a new audio segmentation method based on pith frequency with acceptable accuracy along with a higher computation speed than BIC-based methods. This algorithm is about 2.4 times faster than the BIC-based segmentation an about %1 higher than that of the BIC-based method in accuracy.