Proofpoint Unveils the Industry's Most Advanced Anti-Spam Laboratory.

Business Editors/High-Tech Writers

CUPERTINO, Calif.--(BUSINESS WIRE)--Oct. 28, 2003

World-Class Team of Scientists and Researchers Dedicated to

Turning the Tide of Spam Through the Most Scientifically

Powerful Detection Techniques

Proofpoint, Inc., the leader in large enterprise anti-spam solutions, today unveiled its Anti-Spam Laboratory, the world's most advanced center for spam research and analysis, and the first to be based on advanced Machine Learning science. Providing the research and development behind Proofpoint's groundbreaking MLX(TM) Machine Learning technology, the Proofpoint Anti-Spam Laboratory is dedicated to advancing spam detection techniques and ending the spam deluge through scientific research and applied development by its team of scholars, researchers and scientists.

"Spam attacks continually increase in volume and sophistication, and only the most advanced techniques will be successful in fighting this epidemic. The Proofpoint Anti-spam Laboratory provides the industry a much needed scientific discipline that has been missing up until now, allowing spammers to carry on terrorizing corporate inboxes," said Gary Steele, president and CEO of Proofpoint. "Today we also released our Proofpoint Protection Server 1.5 with new MLX technology, taking a big step toward stopping spam, and creating the basis for message classification required in helping corporations adhere to increasingly stringent email regulations."

MLX technology is a revolutionary system that provides the world's most advanced anti-spam solution by combining:

-- Recent advances in the science of Machine Learning, such as

Support Vector Machines

-- Proven anti-spam methods: Bayesian techniques, heuristics, and

message characteristic analysis

-- An ongoing R&D process that drives Proofpoint's Dynamic Update

Service(TM) to enterprise-based Proofpoint server, keeping

them ahead of new spam techniques

Proofpoint has assembled a world-class team of scientists for the Anti-Spam Laboratory, unparalleled in its cross-disciplinary depth and breadth. The team consists of researchers and engineers with deep roots across several relevant disciplines including Machine Learning, statistics, natural language processing, information classification, messaging and security. The laboratory team is bolstered by specific direction and guidance from Proofpoint's Technical Advisory Council, whose members are world-renowned technologists and thought leaders from leading institutions in academia and industry.

"The Proofpoint Anti-Spam Laboratory has already pushed the envelope in building our MLX technology for spam detection," stated Eric Hahn, founder and chairman of Proofpoint. "The addition of our newly formed Technical Advisory Council will make a huge difference for us in a variety of ways, as we continue to transform the human-intensive art of content classification into an automatic, scalable scientific discipline that can benefit organizations."

Technical Advisory Council

The Proofpoint Technical Advisory Council provides tactical and strategic guidance on specific areas related to MLX technology, content classification, messaging and security, in addition to further scaling the infrastructure for the Proofpoint Protection Server. Each member of the council brings a different area of expertise, and together, they help ensure that the Proofpoint Anti-Spam Laboratory stays at the forefront of the industry. Members include:

-- David Crocker, inventor of internet mail format (RFC822) & 45

other key internet standards

-- Philip Hallam-Baker, Principal Scientist, VeriSign, Inc.,

leader in Internet security initiatives

-- Tim Howes, CTO Opsware, Inc., co-inventor of LDAP (Internet

standard for directories);

-- Dan Jurafsky, Asst. Professor, Stanford University, recognized

authority in natural language processing, computational

linguistics and probabilistic models for human syntactic

parsing;

-- Hongyuan Zha, Assoc. Professor, Penn. State -- leading

researcher in scientific computing, Machine Learning and

pattern recognition;

-- ChengXiang Zhai, Asst. Professor, U. of I. -- Urbana-Champaign

-- expert in natural language processing, personalization of

information retrieval and biosequence pattern analysis

"I'm delighted to be a member of Proofpoint's world-class Technical Advisory Council," said Phillip Hallam-Baker. "If spam levels continue to grow unabated, they pose a serious security risk to corporate communications. This is the single most important area of research we can undertake, not only because of the tremendous amount of spam that must be dealt with every day, but also because of the dangerous payloads they may contain."

"Spammers are very clever at getting around older first generation anti-spam solutions," said Dan Jurafsky. "Proofpoint is using truly cutting edge computer science to fight spam. I'm pleased to be part of the advisory council to help preserve the usefulness of the Internet as a communications medium."

"Proofpoint's spam corpus, analysis techniques and Machine Learning models, have proven to be highly predictive and statistically valid," said Hongyuan Zha. "Although spammers continue to evolve, we are able to stay a step ahead of them with the application of novel statistical and Machine Learning methods."

Proofpoint MLX(TM) Technology developed by the Proofpoint

Anti-Spam Laboratory

MLX is a revolutionary spam detection system based on the principles of Machine Learning. Developed by Proofpoint's Anti-Spam Laboratory researchers, MLX incorporates a full range of classification methods, from legacy approaches such as heuristics and Bayesian analysis, to state-of-the-art Machine Learning algorithms such as those used in genomic sequencing analysis, including Logistical Regression and Support Vector Machines. Key to MLX's accuracy is the work done at the Proofpoint Anti-Spam Laboratory, enabling MLX's classifiers to evolve as the nature of spam changes.

The researchers use systems based on rigorous scientific algorithms to analyze millions of spam emails. This allows them to distill more than 50,000 essential text and non-text attributes that can be used as indicators of spam and valid email. Once attributes are extracted, the team runs them through classifiers that train against the above-mentioned data to statistically balance the attribute weights. The weights, attributes, and classifiers together provide a robust snapshot of Internet spam, which contains enough information to accurately classify business email as spam or valid. The latest weights, attributes, and classifiers are packaged into regular updates and automatically and securely deployed to each customer's Proofpoint Protection Server.

Using their combined expertise across Machine Learning, natural language processing, security and messaging, Laboratory researchers continue to push the envelope in spam detection techniques, ensuring that MLX spam effectiveness maintains itself over time. Ongoing work by Laboratory researchers continues to explore novel attribute extraction algorithms and new classifier methodologies.

Technical Advisory Council Background

David Crocker

David Crocker brings a thorough understanding of Internet infrastructure and messaging techniques. Crocker has been leading and authoring Internet standards for thirty years, covering Internet mail, instant messaging, facsimile and EDI. He has authored more than 45 IETF Request For Comments, including RFC 822 (Standard for the Format of ARPA Internet Text Messages), which defines the format of current Internet email messages. Crocker worked in the ARPA network research community during the 1970s, and after that led product development efforts at MCI and various Silicon Valley companies. For his work on email, Crocker is a co-recipient of the 2004 IEEE Internet Award.

Phillip Hallam-Baker, Ph.D

Phillip Hallam-Baker brings expertise in security, both inside and outside messaging. He is Principal Scientist and Web Services Security Architect for VeriSign, Inc. He is responsible for driving and delivering key security specifications and technologies through industry recognized standards bodies and other organizations. Hallam-Baker has a degree in Electronic Engineering from Southampton University and a Doctorate in Computer Science from the Nuclear Physics Department at Oxford University. He was appointed a Post Doctoral Research Associate at DESY in 1992 and CERN Fellow in 1993. He is currently a researcher at the World Wide Web Consortium at MIT.

Tim Howes, Ph.D.

Tim Howes co-invented the Lightweight Directory Access Protocol (LDAP), the Internet standard for directories. As CTO and EVP of Development at Opsware Inc., Howes is responsible for product development and strategy. Prior to co-founding Opsware Inc., Howes was Vice President of Technology for America Online, Inc., where he worked on technology strategy, acquisitions and investments. Before his work at AOL, he was Chief Technology Officer of Netscape's Server Products division and was Chief Architect of several Netscape server products. Before joining Netscape, Howes was a researcher and National Science Foundation Project Director at the University of Michigan. Howes has authored or co-authored dozens of IETF Requests for Comments, including RFC 1777 (Lightweight Directory Access Protocol).

Dan Jurafsky, Ph.D.

Dan Jurafsky brings expertise in combining computational linguistic analysis with probabilistic modeling. Jurafsky is a new Associate Professor at Stanford University, department of Linguistics. His research focuses on integrating sophisticated linguistic knowledge (of phonology, syntax, semantics, and discourse) with probabilistic, evidential, and information-theoretic algorithms. His work in cognitive modeling includes one of the first probabilistic models for human syntactic parsing, and recent work focuses on probabilistic models of pronunciation in speech, sentence-understanding, and question-answering.

Hongyuan Zha, Ph.D.

Hongyuan Zha brings expertise in advanced Machine Learning theories and practical applications. Zha currently serves as Associate Professor at The Pennsylvania State University's Department of Computer Science and Engineering. As a member of the Scientific Computing Group, Zha focuses on computing with applications in data mining, Machine Learning, pattern recognition and information retrieval.

ChengXiang Zhai, Ph.D.

ChengXiang Zhai brings expertise in the application of statistical language models to the management of textual information. Zhai is an Assistant Professor at The University of Illinois at Urbana-Champaign's Department of Computer Science. As a member of the Database and Information Sciences Group, Zhai focuses on information retrieval, natural language processing and bioinformatics.

About Proofpoint, Inc.

Proofpoint provides business software for large enterprises to stop spam, protect against email viruses and to ensure compliance with corporate policies and regulations through their messaging infrastructure. The company's flagship product, Proofpoint Protection Server is a software breakthrough that employs Proofpoint's MLX technology, an advanced Machine Learning system developed at Proofpoint's Anti-spam Laboratory to govern the flow of digital correspondence. Proofpoint was founded by technology visionary and former CTO of Netscape Communications, Eric Hahn. The Cupertino, California-based company is funded by Benchmark Capital, Inventures Group, Mohr, Davidow Ventures, RRE Ventures and Stanford University. For more information, please visit www.proofpoint.com.

Note: Proofpoint, Inc., Proofpoint Protection Server(TM), MLX(TM) and Dynamic Update Service(TM) are registered trademarks of Proofpoint, Inc. All other trademarks are property of their respective holders.

Related news: Proofpoint Protection Server 1.5 with MLX(TM) (www.proofpoint.com/news/pps15) & Executive Business Advisory Board (www.proofpoint.com/news/eab).