Saleem Karmoshi

Multi-Tenant Data Centers (MTDCs) allo-cate resources to tenants in terms of processors, memory, and storage. However, equal allocation of network resources is often overlooked, leading to unpredictable application performance. To address this issue, we propose Tenant-Aware Resource Allocation (TARA), a virtual resource allocation mechanism for MTDCs. TARA allocates tenants’ virtual net-work resources as virtual ports on the substrate physical net-work, enabling control and management by dedicated con-trollers. In this paper, we introduce a classification method for virtual nodes within Virtual Data Centers (VDCs) aimed at ensuring optimal network performance based on tenant demands. Furthermore, we present a source routing mechanism that utilizes path tables to minimize traffic forwarding delays and enhance network workload efficiency. The TARA model optimizes virtual resource allocation, enhances net-work performance, and simplifies virtual network resource management. Experimental evaluations demonstrate the ef-fectiveness of the TARA system in improving network performance and meeting tenants’ quality of service require-ments.

High-utility itemset mining (HUIM), which is the detection of high-utility itemsets (HUIs) in a transactional database, provides the decision maker with greater flexibility to exploit item utilities, such as quantity and profits, to extract remarkable and efficient database patterns. However, most prevailing empirical articles have focused on HUIs. Nevertheless, in many practical situations, low-utility itemsets (LUIs) maintain a high level of significance and usage (e.g., in security systems and the low-utility itemsets represent the security system vulnerabilities that need monitoring). Hence, this paper proposes a new association rule mining (ARM) framework named low-utility itemset mining (LUIM) that extracts LUIs. Enhancing the performance of LUIM, the researchers here propose innovative HUI generators, determining the generators based on the itemset transaction weight utility (TWU) rather than the support values used in HUG-Miner and GHUI-Miner. Moreover, this paper offers two efficient algorithms called LUG-Miner and LUIMA. The LUG-Miner extracts high and low-utility generators while LUIMA extracts low-utility itemsets using low-utility generators (LUGs). The experimental results on both dense and sparse datasets illuminated the recommended framework, and the algorithms are efficiently operational.

Allocating bandwidth guarantees to applications in the cloud has become increasingly demanding and essential as applications compete to share cloud network resources. However, cloud-computing providers offer no network bandwidth guarantees in a cloud environment, predictably preventing tenants from running their applications. Existing schemes offer tenants practical cluster abstraction solutions emulating underlying physical network resources, proving impractical; however, providing virtual network abstractions has remained an essential step in the right direction. In this paper, we consider the requirements for enabling the application-aware network with bandwidth guarantees in a Virtual Data Center (VDC). We design GANA-VDC, a network virtualization framework supporting VDC application-aware networking with bandwidth guarantees in a cloud datacenter. GANA-VDC achieves scalability using an interceptor to translate OpenFlow features to prompt fine-grained Quality of Service (QoS). Facilitating the expression of diverse network resource demands, we also propose a new Virtual Network (VN) to Physical Network (PN) mapping approach, Graph Abstraction Network Architecture (GANA), which we innovatively introduce in this paper, allowing tenants to provide applications with cloud networking environment, thereby increasing the preservation performance. Our results show GANA-VDC can provide bandwidth guarantee and achieve low time complexity, yielding higher network utility.

Opportunistic Routing (OR) is adapted to improve the performance of low Duty-cycled Wireless Sensor Networks by exploiting its broadcast nature. In contrast to traditional routing, where packets are transmitted along pre-determined paths, OR uses a prioritization metric to select a set of candidates as potential forwarders. This solves the sender's waiting time problem. However, too many candidates may simultaneously wake-up, generating more duplicate packets, occupying the restricted resources and hinder the packet delivery performance. Consciously, to restrict the number of candidates and to counterbalance between the waiting time problem and the duplicate packets problem, this paper proposed a new protocol that combines two main parts. First, each node defines a Candidates Zone (CZ) by a regular geometric shape of four corners. The packets generated by the node will be routed via any path within the CZ. Expressly, the nodes within the CZ are allowed to be selected as candidates. The size of CZ is controlled by the network density. Second, the candidates within the CZ are prioritized based on the OR metric, which is defined as the multiplication of four-distributions: direction distribution, transmission-distance distribution, perpendicular-distance distribution and residual energy distribution. Through an extensive performance evaluation study and simulation of large-scale scenarios, the results demonstrated that our protocol achieved better performance compared to the state-of-the-art solutions in terms of network lifetime, energy consumption, routing efficiency, sender waiting time and duplicate packets.

With the popularity of surveillance video, face detection in surveillance video has become a popular and important topic. Face detection in surveillance video plays an important role in many popular applications such as: personal identification, crowd analysis, database establishment, and abnormal event detection. This paper proposes an unconstrained face detection method for surveillance video, which is not influenced by factors such as face location, expression, posture, scale, and lighting conditions. First, the detection area is initially extracted from the video frame using the improved foreground extraction and skin color detection. Next, we then use the multi-scale sliding window and the cascaded Convolutional Neural Network (CNN) designed in this paper to detect faces. This cascaded network consists of two CNN networks: the first network filters out most of the background area while ensuring the running speed of the whole system and the recall rate of the face, while the second network guarantees the accuracy of the overall system. Finally, we set up a database for the experiment which contained samples from the actual surveillance video. The results of our experiment suggest that the proposed method can obtain good results on unconstrained face detection in surveillance video and can also achieve satisfactory detection speed.

Face images captured by real-world video surveillance applications usually have low resolution. This leads to poor performance or even failure of most face recognition algorithms. As a consequence, identifying the face of the query in low resolution, based on the high-resolution image gallery, proves to be a huge challenge. To address this problem, a novel multi-resolution convolutional neural network (MRCNN) model is proposed in order to study the consistent feature representation from high-resolution and low-resolution face images. First, the corresponding labeled multi-resolution face images are utilized to train the MRCNN model. After this process, the trained model is used as the feature extractor in order to obtain features for the targets in the gallery and query images, respectively. Finally, the nearest neighbor method is applied as the classifier for the purpose of final identification. The experimental results from the two publicly available databases demonstrate the superiority of the proposed MRCNN.