Ndumiyana, David (2013) Hidden markov models and artificial neural networks for spam detection.
Full text not available from this repository.Abstract
The stampede by both individuals and business organizations to the Internet, the acceptance of the Internet as a strategic tool for commerce, sharing of information through communication, electronic surveys and research, entertainment and the exponential growth of the World Wide Web have all combined to rekindle the threat from email spam. The Internet has virtually removed communication barriers between the corporate world and the rest of electronic world due to its ability to share and transmits information within the shortest time possible. Corporations today deploy spam filtering systems to guard the door from the outside world against invasion by unwanted email spam into their email inboxes thereby reducing the impact of junk emails. This paper presents a spam filtering system using hidden markov models and artificial neural networks to filter out spam where word obfuscation on the keyword is conducted to evade detection. To detect spam with word obfuscation on the keywords, we experimented with the use of hidden Markov models (HMMs) to capture the statistical properties of spam variants belonging to the same class. The results showed that our model was able to detect over 90% of spam with a false positive rate of less than 13%. The use of artificial neural network enhanced performance measurement of our filtering system especially on the ability of the system to learn more from any new spam messages that entered the system.
| Item Type: | Article |
|---|---|
| Uncontrolled Keywords: | Hidden markov models, spam keyword obfuscation, spam filter, artificial neural networks. |
| Divisions: | Universities > State Universities > Bindura University of Science Education |
| Depositing User: | Mr. Edmore Sibanda |
| Date Deposited: | 14 May 2018 09:04 |
| Last Modified: | 14 May 2018 09:04 |
| URI: | http://researchdatabase.ac.zw/id/eprint/6302 |
Actions (login required)
 |
View Item |
