UNIFEI - Campus 1: Itajubá IESTI - Instituto de Engenharia de Sistemas e Tecnologia da Informação Artigos Publicados em Periódicos
Use este identificador para citar ou linkar para este item: https://repositorio.unifei.edu.br/jspui/handle/123456789/119
Tipo: Relatório Técnico
Título: An anti-spam prototype.
Autor(es): CARPINTEIRO, Otávio Augusto Salgado
ALMEIDA, Diego P.
MOREIRA, Edmilson M.
Abstract: Anti-spam systems are software systems which filter spam e-mails. Spam e-mails are nowadays a serious problem which causes high losses to the institutions. This report proposes a new anti-spam prototype made up by three stages. The first, the pre-processing, analyses the e-mails to search for known spam patterns, as well as performs eliminations or replacements to simplify the e-mails and to make them uniform. The second stage, the feature selection, identifies the most relevant features of the e-mails. The third stage, the classification, consists in an artificial neural model — the multilayer perceptron — to classify the e-mails. The anti-spam prototype is exhaustively tested on three public corpora — SpamAssassin, LingSpam and Trec 2007 — available in the Internet. The prototype performance is assessed according to the percentage of correct classifications in both e-mail classes — legitimate (ham) and spam. It is also assessed the time spent in training and testing of the neural model. The results obtained are very promising. The anti-spam prototype has very good performance on the three corpora.
Citação: CARPINTEIRO, Otávio A. S.; ALMEIDA, Diego P.; MOREIRA, Edmilson M. An anti-spam prototype. Technical Report IESTI 001, Research Group on Systems and Computer Engineering, Federal University of Itajubá, 2015.
URI: https://repositorio.unifei.edu.br/jspui/handle/123456789/119
Data do documento: Mai-2015
Aparece nas coleções:Artigos Publicados em Periódicos

Arquivos associados a este item:
Arquivo Descrição TamanhoFormato 
carpinteiro_artigo_iesti-002.pdf369,8 kBAdobe PDFVisualizar/Abrir


Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.