No abstract

Face and iris are very common individual bio-metric features for person identification. Face recognition is the method of identification a person uniquely using face. Principal component analysis is one of the algorithms for face recognition. Iris recognition in another method of person identification using iris. Very popular iris recognition method is Daugman algorithm. Unimodal biometric system has various difficulties to detect a person like noisy and unusual data. Multimodal biometric system combined more than one individual modalities like face and iris to increase the efficiency. In this work, we combined principal component analysis and Daugman algorithm with ORL, YALE, CASIA and Real face dataset to combine face and iris recognition to improve the recognition efficiency.

Abstract-Security has become a critical factor in today's computation systems. The security threats that risk our confidential information can come in form of seemingly legitimate client request to server. While illegitimate requests consume the number of connections a server can handle, no valid new connections can be made. This scenario, named SYNflooding attacks can be controlled through a fair scheduling algorithm that provides more opportunity to legal requests. This paper proposes a detailed scheduling approach named Largest Processing Time Rejection-Particle Swarm Optimization (LPTR-PSO) that defends the server against varying intensity SYN-flood attack scenarios through a three-phased algorithm. This novel approach considers the number of half-open connections in the server buffer and chooses a phase accordingly. The simulation results show that the proposed defense strategy improves the performance of under attack system in terms of memory occupancy of legal requests and residence time of attack requests.

No abstract

An enormous number of world populations in current time are unique in that sense that they have no broad language because of the absence of their hearing capability. The people with hearing impairment have their own language called Sign Language however it is hard for understanding to general individuals [1]. Sign digits are additionally a significant piece of gesture based communication. So a machine interpreter is important to permit them to speak with general individuals. For making their language justifiable to general individual’s computer vision based arrangements are notable these days. In this exploration of work we target to develop a model based on CNN to deal with the recognition of Sign Language digits. A dataset of 10 classes is used to train (70%), validation (20%) and test (10%) of the network. We consider three different models of CNN network to train and test the accuracy of sign digit. Among the three model transfer learning based pre-trained CNN performs better with test accuracy of 92%.