View in EDS

SPAM FILTERING THROUGH NEURAL NETWORK.

Saved in:
Bibliographic Details
Title: SPAM FILTERING THROUGH NEURAL NETWORK.
Authors: Atanasova, Todorka, Parusheva, Silvia, Kostadinova, Elena
Source: Proceedings of the International Multidisciplinary Scientific GeoConference SGEM; 2016, Vol. 1, p383-388, 6p
Subject Terms: SPAM filtering (Email), ARTIFICIAL neural networks, MACHINE learning, ALGORITHMS, QUASI-Newton methods
Abstract: The changing characteristics of email spam messages together with the inefficient spam filters require new approaches for spam reduction. One of them is using machine based system learning - a neural network. This approach is used for recognition of spam characteristics and spam correct classification. The ability of neural networks to "learn" from examples makes them highly adaptable and powerful. 1600 messages (spam and not spam) in English, received in a period of several months, have been used in the current research for the training of a neural network of type multilayer perceptron. A system, which does preliminary processing of email messages, is created. The language used to create the system is Java. The platform Eclipse IDE for Java Developers is used. Alyuda NeuroIntelligence is used to create the neural network. The developed system for preliminary processing downloads the messages directly from the mail box through Internet Message Access Protocol. This is done through JavaMail API. The list of downloaded messages is processed according to certain characteristics. A vector with the received results is generated for each processed email, after which all vectors are recorded in CSV file. This file is uploaded to the software for neural networks Alyuda NeuroIntelligence. A multilayer perceptron is created (15-12-1). It is trained through the algorithm Limited Memory Quasi-Newton. Experiments made show that trained network in such way, classify correctly the spam and a mistake has been made in only 50 cases out of total 1200. Therefore a filtering spam system with the use of neural network can be built on the base of the descriptive characteristics of the email messages. [ABSTRACT FROM AUTHOR]
Copyright of Proceedings of the International Multidisciplinary Scientific GeoConference SGEM is the property of STEF92 Technology Ltd. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Complementary Index
Description
Abstract:The changing characteristics of email spam messages together with the inefficient spam filters require new approaches for spam reduction. One of them is using machine based system learning - a neural network. This approach is used for recognition of spam characteristics and spam correct classification. The ability of neural networks to "learn" from examples makes them highly adaptable and powerful. 1600 messages (spam and not spam) in English, received in a period of several months, have been used in the current research for the training of a neural network of type multilayer perceptron. A system, which does preliminary processing of email messages, is created. The language used to create the system is Java. The platform Eclipse IDE for Java Developers is used. Alyuda NeuroIntelligence is used to create the neural network. The developed system for preliminary processing downloads the messages directly from the mail box through Internet Message Access Protocol. This is done through JavaMail API. The list of downloaded messages is processed according to certain characteristics. A vector with the received results is generated for each processed email, after which all vectors are recorded in CSV file. This file is uploaded to the software for neural networks Alyuda NeuroIntelligence. A multilayer perceptron is created (15-12-1). It is trained through the algorithm Limited Memory Quasi-Newton. Experiments made show that trained network in such way, classify correctly the spam and a mistake has been made in only 50 cases out of total 1200. Therefore a filtering spam system with the use of neural network can be built on the base of the descriptive characteristics of the email messages. [ABSTRACT FROM AUTHOR]
ISSN:13142704