DSPAM 3.8.0

DSPAM is a server-side statistical anti-spam agent for Unix email servers.
DSPAM masquerades as the email servers local delivery agent and effectively filters spam using a combination of de-obfuscation techniques, specialized algorithms, and statistical analysis.
The result is an administratively maintenance-free, self-learning anti-spam tool. DSPAM has yielded real-world success rates beyond 99.9% accuracy with less than a 0.01% chance of false positives.
The DSPAM project attempts to set itself apart from other filters by focusing on the following areas:
- DSPAM has a strong drive for research. Many new algorithms and approaches to fighting spam have come out of the DSPAM project. Some of the approaches deployed in DSPAM include Concept Identification, Neural Networking, Message Inoculation , advanced de-obfuscation techniques, and a new noise reduction algorithm called Bayesian Noise Reduction. Were always looking for new approaches to improving the accuracy of DSPAM.
- A strong focus on large-scale implementation support. The largest implementation of DSPAM weve heard about to-date involves 350,000 users, with the next largest being around 125,000, then 100,000. DSPAM has been designed to run with a very short execution time (between 0.01s - 0.03s real time for classification and between 0.03s - 0.10s real time for training, on average hardware), and has been equipped with a storage driver API allowing several different storage mechanisms to be used. Depending on disk space constraints, accuracy can be traded off for additional disk space or vice-versa.
- Usability. DSPAM was designed with "grandma" in mind. Users can retrain by either forwarding any spam they receive to a spam address, or (in v3.4.2+) use the history function of the included web interface to quickly mark spam and deliver false positives. End-users dont need to know any commandline utilities or other complexities plaguing some other such tools. Functions such as whitelisting and keyword inventory are automatic (based on statistical functions) and therefore require no user intervention.
Main features:
- System-wide administratively-maintenance free filtering. The DSPAM agent can integrate into just about any network and can even be implemented as an SMTP gateway.
- A simple-to-use learning mechanism. DSPAM allows users to simply forward their spam to their "spam email address" for learning, eliminating any learning curve necessary to make it usable by your customers. The information used in every calculation is temporarily stored on the server, enabling DSPAM to relearn the original message by looking for a small signature in the forwarded spam. As a result, users dont have to be trained to bounce messages around, and administrators dont have to worry about incompatible mail clients.
- Support for a variety of storage implementations. DSPAMs storage driver API allows the administrator to choose how they wish to store data. Currently supported drivers include SQLite, Berkeley DB3, Berkeley DB4, MySQL, PostgreSQL and Oracle.
- Written in C for speed, performance, and scalability. Unlike Python or PERL solutions DSPAM is written in a low-level compiled language, meaning there is very little overhead. DSPAM runs fast, efficient, and doesnt depend on any third-party language interpreters.
- MTA support. DSPAM works great with Sendmail, Postfix, Qmail, Courier, and Exim, and should work well with many other MTAs. In the event you happen to run something like Exchange, DSPAM can be implemented on your network as an SMTP gateway. Just point your MX at it and configure it to relay to your mail server.
Enhancements:
- This is an unofficial release, but significant.
- Several significant bugfixes have been made.
- Bill Yerazunis Orthogonal Sparse Bigram (OSB) tokenizer algorithm has been added.
- The code has been significantly optimized to run much faster and with fewer resources.