| Publisher | University of Ioannina | ||
|---|---|---|---|
| Format | 217.8KB PDF | Date added | 06 Oct 2008 |
| Topics | Network Security, Security Management, Spam - E-mail Fraud - Phishing | ||
| Downloads | 0 | ||
This paper proposes a novel approach for spam filtering based on the use of Deep Belief Networks (DBNs). In contrast to conventional feedfoward neural networks having one or two hidden layers, DBNs are feedforward neural networks with many hidden layers. Until recently it was not clear how to initialize the weights of deep neural networks, which resulted in poor solutions with low generalization capabilities. A greedy layer-wise unsupervised algorithm was recently proposed to tackle this problem with successful results. This paper presents a methodology for spam detection based on DBNs and evaluates its performance on three widely used datasets. They also compare their method to Support Vector Machines (SVMs) which is the state-of-the-art method for spam filtering in terms of classification performance.
Did you find this white paper useful?
Related white papers
MessageLabs Intelligence : 2009 security Predictions
Having analyzed the global threat landscape for almost a decade, MessageLabs Team Skeptic™ is comprised of many world-renowned malware and spam experts who have a global view of threats across...
IDC Vendor Spotlight
Organised ubiquity is a must for organisations to sucessfully "project" their users in any given landspace, at any given time, with secuirty policy. This White Paper covers issues surrounding secure...
Trend Micro Enterprise Security white paper
This white paper reviews the content security threat landscape and how it has evolved into a more dangerous and high risk environment. The paper discussed how conventional content security approaches...
Smart Protection E-Book
"Outthink the Threat: Why conventional protection is no match for new data-stealing malware & the Trend Micro Smart Protection Network is." Discover how cyber criminals are rendering traditional security solutions...
Secure Desktop On-Demand Webcast
The desktop or endpoint is one of the most vulnerable parts of your environment. Threats are everywhere. You have users who love to experiment with device settings (only to wonder...
Trend Micro Web Threats WP
Web threats pose a broad range of risks, including financial damages, identity theft, loss of confidential business information, theft of network resources, damaged brand or personal reputation, and erosion of...
The Past, Present and Future of Whitelisting
The explosion in malware and zero-day vulnerabilities over the last several years has limited the effectiveness of blacklisting technology. While still important, it must be augmented with other technologies such...
