This thesis solves possibility of using entropy for anomaly detection in data communication and especially for security attacks. The main advantage of using entropy is ability to identify unknown attacks because entropy detects changes in network traffic but not the content as existing methods. In this work was tested the suitability of different models entropy (Shannon, Renyi, Tsallis). Also been tested the effect of Renyi and Tsallis parameter on resulting entropy. From the resulting values, I found that all tested entropy achieve good result in the identification of anomalies in network traffic.

Theme of the Master's thesis is issue of backup power units and its attributes. There is also complete proposal of UPS power unit which can be used to control and monitor backup batteries. Power supply is fully controlled via Ethernet interface using TCI/IP protocol. The other chapter describes complete development of TCI/IP GSM unit which can be used for communication interface between GSM and Ethernet protocol. UPS and GSM modules are developed in details. The Master.s thesis includes all needed information for HW unit development and also for controlling software as well. Universal solution of the modules allows number possibilities of further development.

Mémoire numérisé par la Direction des bibliothèques de l'Université de Montréal.

No description available.

No description available.

Asymmetric digital subscriber line (ADSL) is a broadband access technology capable of delivering large bandwidth over existing copper telephone line infrastructure. This research aims at characterizing and analyzing TCP/IP performance in presence of a new protocol stack (TCP/IP over PPP and ATM) being promoted for one of the ADSL network architectures. Using extensive simulations, we verify the adverse effects of asymmetric links on the performance of TCP and additional throughput degradation caused by the overhead at the AAL5- ATM layers. This study involves unidirectional as well as bi-directional data transfer using different traffic mixes including bursty and non-bursty types of traffic. Bi-directional data transfer over asymmetric links results in ACK compression wherein TCP acknowledgements (ACKs) get bunched together behind larger data packets, further exacerbating the effect of asymmetry on TCP performance. By implementing the simulation model for PPP encapsulation over AAL5, we characterize its effect in terms of throughput degradation and excessive delay. We quantify the improvement in the throughput obtained by delaying the TCP ACKs and by TCP/IP header compression. These techniques being effective for unidirectional traffic over asymmetric links, however, do not prove as effective when ATM enters the scenario or in presence of bi-directional data transfer. Further, we implemented a simulation model of the Smart ACK Dropper (SAD), a technique to regulate the flow of TCP ACKs. Considerable improvement in performance especially in the presence of unidirectional data transfer is achieved using the SAD technique. Although the improvement is to a lesser extent in the presence of bi-directional data traffic, SAD helps the network in quickly recovering from the impact of ACK compression. We also propose and implement certain customized queuing/scheduling and policing mechanisms to enable differentiated servicing of TCP ACKs and data packets, and mitigate the effect of ACK compression. While providing considerable TCP performance improvement in presence of SAD, custom queuing also allows fair sharing of bandwidth between TCP flows, unlike priority queuing, which starves the low priority flow. The committed access rate (CAR)policing scheme provides considerable performance improvement when used with SAD, and is especially useful when TCP ACKs compete with bursty data traffic over the slower upstream. / Master of Science

Com o recente desenvolvimento de tecnologias de redes de altas taxas de transmissão, tais como Asynchronous Transfer Mode (ATM), o problema da carência por largura de banda foi solucionado. A questão atual é a implementação de sistemas que suportem os protocolos ATM de forma nativa e integral. Atualmente tem-se utilizado aplicativos tradicionais baseados nos protocolos TCP(UDP)/IP no topo da pilha de protocolos ATM. Tal modelo traz redundâncias que implicam diretamente em aumento de overhead na comunicação. Muitos modelos têm sido desenvolvidos para levar de forma direta a aplicação os serviços ATM. O propósito deste trabalho foi comparar o desempenho de transmissões de dados utilizando os protocolos TCP(UDP)/IP sobre ATM com transmissões de dados no modo ATM nativo. Para tal, utilizamos a plataforma de pesquisa HARP do sistema operacional FreeBSD. Tal plataforma implementa o modelo Classical IP sobre ATM utilizando os serviços AAL5 e ainda fornece uma API para sockets ATM permitindo que uma aplicação tenha acesso direto a camada AALS. Testes de taxa de transmissão, perda de células, e atraso na rede, foram realizados com base na modificação de parâmetros relacionados à aplicação e sistema operacional, tais como tamanho da mensagem sendo transmitida e tamanho dos buffers de socket. / The solution for the lack of bandwidth was solved by the recent development of high speed networks technologies, such as Asynchronous Transfer Mode - ATM. The question now is the implementation of systems, which provide support for ATM protocols in an integral way. Nowadays, legacy network applications, based in TCP/IP protocols have been used on top of ATM protocol stack. Such approach generates redundances that have as consequence an increase in the communication overhead. Many models have been developed to provide ATM services directly to applications. The goal of this study was to evaluate the performance of data transfers comparing the TCP/IP protocols over ATM with the performance of data transfers in native ATM mode. The Host ATM Research Plataform (HARP) was utilized on the FreeBSD operating system. This plataform implements Classical IP over ATM, utilizing AAL5 services. Furthermore, it provides an API for ATM sockets, allowing that an application directly access the AAL5 services. Measurements of throughput and investigations about cell loss and delay were carried out altering parameters related to the application and operating system. The parameters adopted were message size and socket buffer sizes.

Made available in DSpace on 2016-04-04T18:31:20Z (GMT). No. of bitstreams: 1 Alberto Lotito.pdf: 2153365 bytes, checksum: 3433c9bb5371bcce62f119c4593fe360 (MD5) Previous issue date: 2007-12-05 / This work intends to perform end-to-end traffic engineering through simulations in broadband multiprotocol networks and evaluate used parameters. We ve proposed and tested parameters used by the network administrator to dimension the action of an algorithm for Traffic Engineering. We dedicated special attention to IP networks interconnected to MPLS networks. Through simulations, we ve addressed questions of performance optimization, evaluating the network packet losses and compare the results with packet loss in the traditional OSPF network achieving improvements of up to 50%. This work contributes with improvements in traffic performance for real world networks, developing practical application since the traditional IP networks up to the present MPLS networks and also for mixed technologies. / Este trabalho tem como objetivos realizar simula??es de engenharia de tr?fego fim a fim em redes que trabalham com protocolos distintos e avaliar par?metros usados. Propusemos e testamos par?metros que servem para que o operador da rede dimensione a atua??o de um algoritmo para engenharia de tr?fego. Demos especial aten??o ?s redes que trabalham com IP interconectando-se a redes MPLS. Executando diversas simula??es, trabalhamos quest?es de melhoria de desempenho por meio de avalia??o da perda de pacotes nas redes testadas comparando-se os resultados com as tradicionais redes com protocolo OSPF e chegando a melhoria de at? 50% se comparado com este. Este trabalho contribui com melhorias em engenharia de tr?fego em redes do mundo real, desenvolvendo aplica??o pr?tica desde as tradicionais redes IP at? as atuais redes MPLS e tamb?m em redes de tecnologias mistas.

Com o recente desenvolvimento de tecnologias de redes de altas taxas de transmissão, tais como Asynchronous Transfer Mode (ATM), o problema da carência por largura de banda foi solucionado. A questão atual é a implementação de sistemas que suportem os protocolos ATM de forma nativa e integral. Atualmente tem-se utilizado aplicativos tradicionais baseados nos protocolos TCP(UDP)/IP no topo da pilha de protocolos ATM. Tal modelo traz redundâncias que implicam diretamente em aumento de overhead na comunicação. Muitos modelos têm sido desenvolvidos para levar de forma direta a aplicação os serviços ATM. O propósito deste trabalho foi comparar o desempenho de transmissões de dados utilizando os protocolos TCP(UDP)/IP sobre ATM com transmissões de dados no modo ATM nativo. Para tal, utilizamos a plataforma de pesquisa HARP do sistema operacional FreeBSD. Tal plataforma implementa o modelo Classical IP sobre ATM utilizando os serviços AAL5 e ainda fornece uma API para sockets ATM permitindo que uma aplicação tenha acesso direto a camada AALS. Testes de taxa de transmissão, perda de células, e atraso na rede, foram realizados com base na modificação de parâmetros relacionados à aplicação e sistema operacional, tais como tamanho da mensagem sendo transmitida e tamanho dos buffers de socket. / The solution for the lack of bandwidth was solved by the recent development of high speed networks technologies, such as Asynchronous Transfer Mode - ATM. The question now is the implementation of systems, which provide support for ATM protocols in an integral way. Nowadays, legacy network applications, based in TCP/IP protocols have been used on top of ATM protocol stack. Such approach generates redundances that have as consequence an increase in the communication overhead. Many models have been developed to provide ATM services directly to applications. The goal of this study was to evaluate the performance of data transfers comparing the TCP/IP protocols over ATM with the performance of data transfers in native ATM mode. The Host ATM Research Plataform (HARP) was utilized on the FreeBSD operating system. This plataform implements Classical IP over ATM, utilizing AAL5 services. Furthermore, it provides an API for ATM sockets, allowing that an application directly access the AAL5 services. Measurements of throughput and investigations about cell loss and delay were carried out altering parameters related to the application and operating system. The parameters adopted were message size and socket buffer sizes.

O crescimento dos serviços de banda-larga em redes de comunicações móveis tem provocado uma demanda por dados cada vez mais rápidos e de qualidade. A tecnologia de redes móveis chamada LTE (Long Term Evolution) ou quarta geração (4G) surgiu com o objetivo de atender esta demanda por acesso sem fio a serviços, como acesso à Internet, jogos online, VoIP e vídeo conferência. O LTE faz parte das especificações do 3GPP releases 8 e 9, operando numa rede totalmente IP, provendo taxas de transmissão superiores a 100 Mbps (DL), 50 Mbps (UL), baixa latência (10 ms) e compatibilidade com as versões anteriores de redes móveis, 2G (GSM/EDGE) e 3G (UMTS/HSPA). O protocolo TCP desenvolvido para operar em redes cabeadas, apresenta baixo desempenho sobre canais sem fio, como redes móveis celulares, devido principalmente às características de desvanecimento seletivo, sombreamento e às altas taxas de erros provenientes da interface aérea. Como todas as perdas são interpretadas como causadas por congestionamento, o desempenho do protocolo é ruim. O objetivo desta dissertação é avaliar o desempenho de vários tipos de protocolo TCP através de simulações, sob a influência de interferência nos canais entre o terminal móvel (UE User Equipment) e um servidor remoto. Para isto utilizou-se o software NS3 (Network Simulator versão 3) e os protocolos TCP Westwood Plus, New Reno, Reno e Tahoe. Os resultados obtidos nos testes mostram que o protocolo TCP Westwood Plus possui um desempenho melhor que os outros. Os protocolos TCP New Reno e Reno tiveram desempenho muito semelhante devido ao modelo de interferência utilizada ter uma distribuição uniforme e, com isso, a possibilidade de perdas de bits consecutivos é baixa em uma mesma janela de transmissão. O TCP Tahoe, como era de se esperar, apresentou o pior desempenho dentre todos, pois o mesmo não possui o mecanismo de fast recovery e sua janela de congestionamento volta sempre para um segmento após o timeout. Observou-se ainda que o atraso tem grande importância no desempenho dos protocolos TCP, mas até do que a largura de banda dos links de acesso e de backbone, uma vez que, no cenário testado, o gargalo estava presente na interface aérea. As simulações com erros na interface aérea, introduzido com o script de fading (desvanecimento) do NS3, mostraram que o modo RLC AM (com reconhecimento) tem um desempenho melhor para aplicações de transferência de arquivos em ambientes ruidosos do que o modo RLC UM sem reconhecimento. / The growth of broadband services in mobile networks has led to a demand for data with faster and better quality transmissions. The mobile network technology called LTE (Long Term Evolution) or fourth generation (4G) came up with the objective of attending this demand for wireless access to services such as Internet access, online games, VoIP and video conferencing. LTE is part of the specifications of 3GPP Releases 8 and 9 operating in all-IP networks and providing transmission rates above 100 Mbps (DL), 50 Mbps (UL), low latency (10 ms) and compatibility with previous versions of mobile networks, 2G (GSM / EDGE) and 3G (UMTS / HSPA). The TCP protocol designed to operate in wired networks presents poor performance over wireless channels such as mobile cellular networks, due mainly to the characteristics of selective fading, shadowing and high error rates coming from the air interface. As all losses are interpreted as caused by congestion the protocol performance is bad. The objective of this dissertation is to evaluate the performance of several types of the TCP protocols through simulations, under the influence of channel interference between the mobile terminal (UE - User Equipment) and a remote server. For this, the NS3 (Network Simulator version 3) software and the protocols TCP Westwood Plus, New Reno, Reno and Tahoe were used. Results have shown that the TCP Westwood Plus protocol has a better performance than others. The New Reno and Reno TCP protocols had similar performance due to the proposed interference model, which has a uniform distribution and so the possibility of loss of consecutive bits is low on the same transmission window. TCP Tahoe, as expected has shown the worst performance among all because it does not have the fast recovery mechanism and its congestion window keeps coming back to one segment after a timeout. It was also observed that the delay has a greater importance in the performance of TCP when comparing with the bandwidth of the access and backbone links importance, once in the tested scenario the bottleneck was present in the air interface. The simulation performed with noise in the Air Interface, introduced by the NS3 fading script, showed that the RLC AM (acknowledged mode) had a better performance than the RLM UM (Unacknowledged mode).