SpamSieve’s accuracy during the test period was a perfect 100%.SpamSieve correctly identified all 62 spam messages as spam, with no false positives and no false negatives.SpamAsssassin’s accuracy at the client level during the test period was a disappointing 77.4%, but this number does not reflect the high-scoring spam that was automatically deleted by SpamAssassin at the server level.SpamAssassin did not produce any false positives.The remaining 37 were not tagged by SpamAssassin (false negatives).Of the spam messages 25 were correctly tagged as spam by SpamAssassin.38% of all email received by the client was spam.An additional 102 messages were received by the client during the same period, bringing the total count to 164 messages ( spam and ham) in all.62 spam messages made it through to my email client (two accounts: one business and one personal).The following results are from a test period of 24 hours in mid-July 2006:

DNS-based blacklisting is the numerically most important countermeasure, stopping the largest number of messages, and this means that subsequent levels have to defend against fewer spam messages. So as messages pass through the various lines of defense, the number of spam messages getting through to subsequent levels is successively reduced. Automatically deleting high-scoring spam (message immediately deleted before arriving at mailbox).Virus filtering (message immediately deleted before arriving at mailbox).Address migration ('User unknown' returned to spammer).Blocking non-local senders for local-only recipients (delivery aborted with 5.7.1/551 reply code).DNS-based blacklisting (connections rejected with 554 reply code).The order in which the techniques were deployed to the server is:īut the order in which the techniques are actually applied to incoming messages is: As such, the number of messages stopped per day for the other countermeasures is reduced but that does not mean that their efficacy has diminished. For example, DNS-based blacklisting was the last countermeasure to be introduced in the above list, and it blocks spam before it gets to any of the other countermeasures. It is difficult to estimate the number of spam messages that is collectively being blocked by these countermeasures because they operate at different levels and were introduced incrementally. Virus filtering: 47 messsages (see ClamAV log sample).Blocking non-local senders for local-only recipients: 178 messages.

Automatically deleting high-scoring spam: 47 messages.

They are not necessarily statistically valid samples, based on a very small data set collected during June and July of 2006, but they do provide some indication of the amount of spam which is being beaten at the server and never makes it through to the client. The following results show the number of messages blocked (rejected or discarded) at the mail server level in a typical 24 hour period due to the anti-spam technologies and techniques linked to from this page.

Address secrecy (see also Using temporary email aliases to do spam monitoring and damage control)Įfficacy Efficacy of server-side spam blocking.

Client-side spam-filtering with SpamSieve.

Blocking non-local senders for local-only recipients.

Automatically deleting high-scoring spam.

Server-side spam-filtering with SpamAssassin more specifically:.

You should also see log entries from SpamSieveLaunchAgent each time it receives a training command from the plug-in.ġ1/12/13 12:53:06.427 PM Mail: SpamSieve Mail Plug-In: Asking SpamSieve to score message: The radical Paul Ryanġ1/12/13 12:53:06.428 PM Mail: SpamSieve Mail Plug-In: Sending request of length 23909 to ġ1/12/13 12:53:06. In the All Messages log in Console, you should see log entries from Mail (SpamSieve Mail Plug-In) at launch and every time an incoming message is processed or you train a message. If the logging is active, you should see a dialog box in Apple Mail if Mail or AppleScript reports an error while training a message. Subject: Do You Need A Loan? If Yes Contact Us At: ZVVA62v4IhAzZ5N6RABPzw= Message -> SpamSieve -> train as Spam leaves the message in the inbox, and does not change the log. I see the message in the log, marked as “Good”. “Training” the message as spam is not working.

Spamsieve score