Join author John Zdziarski for a look inside the brilliant minds that have conceived clever new ways to fight spam in all its nefarious forms. This landmark title describes, in-depth, how statistical filtering is being used by next-generation spam filters to identify and filter unwanted messages, how spam filtering works and how language classification and machine learning combine to produce remarkably accurate spam filters.

After reading Ending Spam , you'll have a complete understanding of the mathematical approaches used by today's spam filters as well as decoding, tokenization, various algorithms (including Bayesian analysis and Markovian discrimination) and the benefits of using open-source solutions to end spam. Zdziarski interviewed creators of many of the best spam filters and has included their insights in this revealing examination of the anti-spam crusade.

If you're a programmer designing a new spam filter, a network admin implementing a spam-filtering solution, or just someone who's curious about how spam filters work and the tactics spammers use to evade them, Ending Spam will serve as an informative analysis of the war against spammers.

TOC Introduction

PART I: An Introduction to Spam Filtering Chapter 1: The History of Spam Chapter 2: Historical Approaches to Fighting Spam Chapter 3: Language Classification Concepts Chapter 4: Statistical Filtering Fundamentals

PART II: Fundamentals of Statistical Filtering Chapter 5: Decoding: Uncombobulating Messages Chapter 6: Tokenization: The Building Blocks of Spam Chapter 7: The Low-Down Dirty Tricks of Spammers Chapter 8: Data Storage for a Zillion Records Chapter 9: Scaling in Large Environments

PART III: Advanced Concepts of Statistical Filtering Chapter 10: Testing Theory Chapter 11: Concept Identification: Advanced Tokenization Chapter 12: Fifth-Order Markovian Discrimination Chapter 13: Intelligent Feature Set Reduction Chapter 14: Collaborative Algorithms

Appendix: Shining Examples of Filtering

Index

Amazon.com® Reader Reviews (Ranked by Helpfulness)

Average Amazon.com® Rating:  Based on 15 Ratings

ivan's review - 2007-08-08
Reviewer Rating:
There is too much (for me) about marginal matters such as the history of spam and minute details of various methods. I was looking for a clear exposition of the principles of filtering and the corresponding mathematics but this I can't find. The term "decision matrix" is used a lot without being defined.The stuff concerning Bayesian filters on page 76 is quite meaningless. It's all very disappointing.

Outstanding as a text for applied Bayesian stats - 2008-06-25
Reviewer Rating:
This is one of my favorite NLP books because it offers an extremely readable introduction to Bayesian statistics in a very applied context. If you don't have a strong background in statistics and/or text classification, this book is a great way to get an intuitive feel for how Bayesian classifiers work. If you're a developer looking to do some coding, what's explained in the book is easy to translate into code. I recommend this book to upper-level undergrads and graduate students in linguistics who take an applied computational linguistic class I teach.

excellent book - 2007-01-03
Reviewer Rating:
Reading this book was fun. I was doing some research on spam and found this book was exactly what I was looking for. This book covers (almost) all aspects of spam, including the history, the current status, the principles of anti-spam systems, statistical algorithms, case studies, etc. This book is a good start point for understanding spams and means to stop them, although it does not contain a lot of in-depth technical details. I was amazed by the author's style, which was quite energetic and entertaining. This book made my research a pleasant experience. I strongly recommend this book for those who are interested to know how spams came and how we fight them.

Excellent book on spam filter,but the "Bayesian Combination Rule" is not quite correct - 2009-02-20
Reviewer Rating:
I am not a spam expert but an expert on Bayesian. I found this book excellent on spam (history, filters, etc). However, on page 75-76, I couldn't recognize that the Bayesian combination (Paul Graham) formula AB/(AB+(1-A)(1-B)) is related to the Bayes' Theorem P(A|B)=P(B|A)P(A)/P(B). So I went to Paul Graham's website http://www.paulgraham.com/naivebayes.html, where I found that Paul got the formula from http://www.mathpages.com/home/kmath267.htm.

It turns out that the formula is correct only under two stringent conditions: 1) the tokens (the most spamy words) in a spam email are independent (not related); 2) a spam-filter user should have roughly equal number of spam emails and legitimate emails over time. One can go to the links to find more details.

But I still think the formula very usefull and it should be called "Paul Graham's Combination Rule" instead.

Great book! - 2007-01-19
Reviewer Rating:
This book provides the history of spam, so we know how it all started, as well as the reasoning and theories behind the current spam technologies, whithout getting bogged down in minutia. I found this book quick and enjoyable to read. Very informative. Highly suggested if you are a sysAdmin (like me) who has or will build a spam filter, or wants to know how they work and why. Good for programmers as well looking for the theories.

Browse Similar Topics

Top Level Categories:
E-Commerce
Internet/Online
Security

Sub-Categories:
E-Commerce > Security
Internet/Online > E-Mail
Security > E-Commerce