1SA-LEARN(1) User Contributed Perl Documentation SA-LEARN(1)
2
6 sa-learn - train SpamAssassin's Bayesian classifier
7
9 sa-learn [options] [file]...
10 sa-learn [options] --dump [ all | data | magic ]
12 Options:
14 --ham Learn messages as ham (non-spam)
16 --spam Learn messages as spam
17 --forget Forget a message
18 --use-ignores Use bayes_ignore_from and bayes_ignore_to
19 --sync Synchronize the database and the journal if needed
20 --force-expire Force a database sync and expiry run
21 --dbpath <path> Allows commandline override (in bayes_path form)
22 for where to read the Bayes DB from
23 --dump [all|data|magic] Display the contents of the Bayes database
24 Takes optional argument for what to display
25 --regexp <re> For dump only, specifies which tokens to
26 dump based on a regular expression.
27 -f file, --folders=file Read list of files/directories from file
28 --dir Ignored; historical compatibility
29 --file Ignored; historical compatibility
30 --mbox Input sources are in mbox format
31 --mbx Input sources are in mbx format
32 --showdots Show progress using dots
33 --progress Show progress using progress bar
34 --no-sync Skip synchronizing the database and journal
35 after learning
36 -L, --local Operate locally, no network accesses
37 --import Migrate data from older version/non DB_File
38 based databases
39 --clear Wipe out existing database
40 --backup Backup, to STDOUT, existing database
41 --restore <filename> Restore a database from filename
42 -u username, --username=username
43 Override username taken from the runtime
44 environment, used with SQL
45 -C path, --configpath=path, --config-file=path
46 Path to standard configuration dir
47 -p prefs, --prefspath=file, --prefs-file=file
48 Set user preferences file
49 --siteconfigpath=path Path for site configs
50 (default: /etc/mail/spamassassin)
51 --cf='config line' Additional line of configuration
52 -D, --debug [area=n,...] Print debugging messages
53 -V, --version Print version
54 -h, --help Print usage message
55
57 Given a typical selection of your incoming mail classified as spam or
58 ham (non-spam), this tool will feed each mail to SpamAssassin, allowing
59 it to 'learn' what signs are likely to mean spam, and which are likely
60 to mean ham.
61 Simply run this command once for each of your mail folders, and it will
63 ''learn'' from the mail therein.
64 Note that csh-style globbing in the mail folder names is supported; in
66 other words, listing a folder name as "*" will scan every folder that
67 matches. See "Mail::SpamAssassin::ArchiveIterator" for more details.
68 SpamAssassin remembers which mail messages it has learnt already, and
70 will not re-learn those messages again, unless you use the --forget
71 option. Messages learnt as spam will have SpamAssassin markup removed,
72 on the fly.
73 If you make a mistake and scan a mail as ham when it is spam, or vice
75 versa, simply rerun this command with the correct classification, and
76 the mistake will be corrected. SpamAssassin will automatically
77 'forget' the previous indications.
78 Users of "spamd" who wish to perform training remotely, over a network,
80 should investigate the "spamc -L" switch.
81
83 --ham
84 Learn the input message(s) as ham. If you have previously learnt
85 any of the messages as spam, SpamAssassin will forget them first,
86 then re-learn them as ham. Alternatively, if you have previously
87 learnt them as ham, it'll skip them this time around. If the
88 messages have already been filtered through SpamAssassin, the
89 learner will ignore any modifications SpamAssassin may have made.
90 --spam
92 Learn the input message(s) as spam. If you have previously learnt
93 any of the messages as ham, SpamAssassin will forget them first,
94 then re-learn them as spam. Alternatively, if you have previously
95 learnt them as spam, it'll skip them this time around. If the
96 messages have already been filtered through SpamAssassin, the
97 learner will ignore any modifications SpamAssassin may have made.
98 --folders=filename, -f filename
100 sa-learn will read in the list of folders from the specified file,
101 one folder per line in the file. If the folder is prefixed with
102 "ham:type:" or "spam:type:", sa-learn will learn that folder
103 appropriately, otherwise the folders will be assumed to be of the
104 type specified by --ham or --spam.
105 "type" above is optional, but is the same as the standard for
107 ArchiveIterator: mbox, mbx, dir, file, or detect (the default if
108 not specified).
109 --mbox
111 sa-learn will read in the file(s) containing the emails to be
112 learned, and will process them in mbox format (one or more emails
113 per file).
114 --mbx
116 sa-learn will read in the file(s) containing the emails to be
117 learned, and will process them in mbx format (one or more emails
118 per file).
119 --use-ignores
121 Don't learn the message if a from address matches configuration
122 file item "bayes_ignore_from" or a to address matches
123 "bayes_ignore_to". The option might be used when learning from a
124 large file of messages from which the hammy spam messages or spammy
125 ham messages have not been removed.
126 --sync
128 Synchronize the journal and databases. Upon successfully syncing
129 the database with the entries in the journal, the journal file is
130 removed.
131 --force-expire
133 Forces an expiry attempt, regardless of whether it may be necessary
134 or not. Note: This doesn't mean any tokens will actually expire.
135 Please see the EXPIRATION section below.
136 Note: "--force-expire" also causes the journal data to be
138 synchronized into the Bayes databases.
139 --forget
141 Forget a given message previously learnt.
142 --dbpath
144 Allows a commandline override of the bayes_path configuration
145 option.
146 --dump option
148 Display the contents of the Bayes database. Without an option or
149 with the all option, all magic tokens and data tokens will be
150 displayed. magic will only display magic tokens, and data will
151 only display the data tokens.
152 Can also use the --regexp RE option to specify which tokens to
154 display based on a regular expression.
155 --clear
157 Clear an existing Bayes database by removing all traces of the
158 database.
159 WARNING: This is destructive and should be used with care.
161 --backup
163 Performs a dump of the Bayes database in machine/human readable
164 format.
165 The dump will include token and seen data. It is suitable for
167 input back into the --restore command.
168 --restore=filename
170 Performs a restore of the Bayes database defined by filename.
171 WARNING: This is a destructive operation, previous Bayes data will
173 be wiped out.
174 -h, --help
176 Print help message and exit.
177 -u username, --username=username
179 If specified this username will override the username taken from
180 the runtime environment. You can use this option to specify users
181 in a virtual user configuration when using SQL as the Bayes
182 backend.
183 NOTE: This option will not change to the given username, it will
185 only attempt to act on behalf of that user. Because of this you
186 will need to have proper permissions to be able to change files
187 owned by username. In the case of SQL this generally is not a
188 problem.
189 -C path, --configpath=path, --config-file=path
191 Use the specified path for locating the distributed configuration
192 files. Ignore the default directories (usually
193 "/usr/share/spamassassin" or similar).
194 --siteconfigpath=path
196 Use the specified path for locating site-specific configuration
197 files. Ignore the default directories (usually
198 "/etc/mail/spamassassin" or similar).
199 --cf='config line'
201 Add additional lines of configuration directly from the command-
202 line, parsed after the configuration files are read. Multiple
203 --cf arguments can be used, and each will be considered a separate
204 line of configuration.
205 -p prefs, --prefspath=prefs, --prefs-file=prefs
207 Read user score preferences from prefs (usually
208 "$HOME/.spamassassin/user_prefs").
209 --progress
211 Prints a progress bar (to STDERR) showing the current progress. In
212 the case where no valid terminal is found this option will behave
213 very much like the --showdots option.
214 -D [area,...], --debug [area,...]
216 Produce debugging output. If no areas are listed, all debugging
217 information is printed. Diagnostic output can also be enabled for
218 each area individually; area is the area of the code to instrument.
219 For example, to produce diagnostic output on bayes, learn, and dns,
220 use:
221 spamassassin -D bayes,learn,dns
223 For more information about which areas (also known as channels) are
225 available, please see the documentation at:
226 C<http://wiki.apache.org/spamassassin/DebugChannels>
228 Higher priority informational messages that are suitable for
230 logging in normal circumstances are available with an area of
231 "info".
232 --no-sync
234 Skip the slow synchronization step which normally takes place after
235 changing database entries. If you plan to learn from many folders
236 in a batch, or to learn many individual messages one-by-one, it is
237 faster to use this switch and run "sa-learn --sync" once all the
238 folders have been scanned.
239 Clarification: The state of --no-sync overrides the
241 bayes_learn_to_journal configuration option. If not specified, sa-
242 learn will learn to the database directly. If specified, sa-learn
243 will learn to the journal file.
244 Note: --sync and --no-sync can be specified on the same
246 commandline, which is slightly confusing. In this case, the
247 --no-sync option is ignored since there is no learn operation.
248 -L, --local
250 Do not perform any network accesses while learning details about
251 the mail messages. This will speed up the learning process, but
252 may result in a slightly lower accuracy.
253 Note that this is currently ignored, as current versions of
255 SpamAssassin will not perform network access while learning; but
256 future versions may.
257 --import
259 If you previously used SpamAssassin's Bayesian learner without the
260 "DB_File" module installed, it will have created files in other
261 formats, such as "GDBM_File", "NDBM_File", or "SDBM_File". This
262 switch allows you to migrate that old data into the "DB_File"
263 format. It will overwrite any data currently in the "DB_File".
264 Can also be used with the --dbpath path option to specify the
266 location of the Bayes files to use.
267
269 There are now multiple backend storage modules available for storing
270 user's bayesian data. As such you might want to migrate from one
271 backend to another. Here is a simple procedure for migrating from one
272 backend to another.
273 Note that if you have individual user databases you will have to
275 perform a similar procedure for each one of them.
276 sa-learn --sync
278 This will sync any outstanding journal entries
279 sa-learn --backup > backup.txt
281 This will save all your Bayes data to a plain text file.
282 sa-learn --clear
284 This is optional, but good to do to clear out the old database.
285 Repeat!
287 At this point, if you have multiple databases, you should perform
288 the procedure above for each of them. (i.e. each user's database
289 needs to be backed up before continuing.)
290 Switch backends
292 Once you have backed up all databases you can update your
293 configuration for the new database backend. This will involve at
294 least the bayes_store_module config option and may involve some
295 additional config options depending on what is required by the
296 module. (For example, you may need to configure an SQL database.)
297 sa-learn --restore backup.txt
299 Again, you need to do this for every database.
300 If you are migrating to SQL you can make use of the -u <username>
302 option in sa-learn to populate each user's database. Otherwise, you
303 must run sa-learn as the user who database you are restoring.
304
306 (Thanks to Michael Bell for this section!)
307 For a more lengthy description of how this works, go to
309 http://www.paulgraham.com/ and see "A Plan for Spam". It's reasonably
310 readable, even if statistics make me break out in hives.
311 The short semi-inaccurate version: Given training, a spam heuristics
313 engine can take the most "spammy" and "hammy" words and apply
314 probabilistic analysis. Furthermore, once given a basis for the
315 analysis, the engine can continue to learn iteratively by applying both
316 the non-Bayesian and Bayesian rulesets together to create evolving
317 "intelligence".
318 SpamAssassin 2.50 and later supports Bayesian spam analysis, in the
320 form of the BAYES rules. This is a new feature, quite powerful, and is
321 disabled until enough messages have been learnt.
322 The pros of Bayesian spam analysis:
324 Can greatly reduce false positives and false negatives.
326 It learns from your mail, so it is tailored to your unique e-mail
327 flow.
328 Once it starts learning, it can continue to learn from SpamAssassin and
330 improve over time.
331 And the cons:
333 A decent number of messages are required before results are useful for
335 ham/spam determination.
336 It's hard to explain why a message is or isn't marked as spam.
337 i.e.: a straightforward rule, that matches, say, "VIAGRA" is easy
338 to understand. If it generates a false positive or false negative,
339 it is fairly easy to understand why.
340 With Bayesian analysis, it's all probabilities - "because the past
342 says it is likely as this falls into a probabilistic distribution
343 common to past spam in your systems". Tell that to your users!
344 Tell that to the client when he asks "what can I do to change
345 this". (By the way, the answer in this case is "use whitelisting".)
346 It will take disk space and memory.
348 The databases it maintains take quite a lot of resources to store
349 and use.
350
352 Still interested? Ok, here's the guidelines for getting this working.
353 First a high-level overview:
355 Build a significant sample of both ham and spam.
357 I suggest several thousand of each, placed in SPAM and HAM
358 directories or mailboxes. Yes, you MUST hand-sort this - otherwise
359 the results won't be much better than SpamAssassin on its own.
360 Verify the spamminess/haminess of EVERY message. You're urged to
361 avoid using a publicly available corpus (sample) - this must be
362 taken from YOUR mail server, if it is to be statistically useful.
363 Otherwise, the results may be pretty skewed.
364 Use this tool to teach SpamAssassin about these samples, like so:
366 sa-learn --spam /path/to/spam/folder
367 sa-learn --ham /path/to/ham/folder
368 ...
369 Let SpamAssassin proceed, learning stuff. When it finds ham and
371 spam it will add the "interesting tokens" to the database.
372 If you need SpamAssassin to forget about specific messages, use the
374 --forget option.
375 This can be applied to either ham or spam that has run through the
376 sa-learn processes. It's a bit of a hammer, really, lowering the
377 weighting of the specific tokens in that message (only if that
378 message has been processed before).
379 Learning from single messages uses a command like this:
381 sa-learn --ham --no-sync mailmessage
382 This is handy for binding to a key in your mail user agent. It's
384 very fast, as all the time-consuming stuff is deferred until you
385 run with the "--sync" option.
386 Autolearning is enabled by default
388 If you don't have a corpus of mail saved to learn, you can let
389 SpamAssassin automatically learn the mail that you receive. If you
390 are autolearning from scratch, the amount of mail you receive will
391 determine how long until the BAYES_* rules are activated.
392
394 Learning filters require training to be effective. If you don't train
395 them, they won't work. In addition, you need to train them with new
396 messages regularly to keep them up-to-date, or their data will become
397 stale and impact accuracy.
398 You need to train with both spam and ham mails. One type of mail alone
400 will not have any effect.
401 Note that if your mail folders contain things like forwarded spam,
403 discussions of spam-catching rules, etc., this will cause trouble. You
404 should avoid scanning those messages if possible. (An easy way to do
405 this is to move them aside, into a folder which is not scanned.)
406 If the messages you are learning from have already been filtered
408 through SpamAssassin, the learner will compensate for this. In effect,
409 it learns what each message would look like if you had run
410 "spamassassin -d" over it in advance.
411 Another thing to be aware of, is that typically you should aim to train
413 with at least 1000 messages of spam, and 1000 ham messages, if
414 possible. More is better, but anything over about 5000 messages does
415 not improve accuracy significantly in our tests.
416 Be careful that you train from the same source -- for example, if you
418 train on old spam, but new ham mail, then the classifier will think
419 that a mail with an old date stamp is likely to be spam.
420 It's also worth noting that training with a very small quantity of ham,
422 will produce atrocious results. You should aim to train with at least
423 the same amount (or more if possible!) of ham data than spam.
424 On an on-going basis, it is best to keep training the filter to make
426 sure it has fresh data to work from. There are various ways to do
427 this:
428 1. Supervised learning
430 This means keeping a copy of all or most of your mail, separated
431 into spam and ham piles, and periodically re-training using those.
432 It produces the best results, but requires more work from you, the
433 user.
434 (An easy way to do this, by the way, is to create a new folder for
436 'deleted' messages, and instead of deleting them from other
437 folders, simply move them in there instead. Then keep all spam in
438 a separate folder and never delete it. As long as you remember to
439 move misclassified mails into the correct folder set, it is easy
440 enough to keep up to date.)
441 2. Unsupervised learning from Bayesian classification
443 Another way to train is to chain the results of the Bayesian
444 classifier back into the training, so it reinforces its own
445 decisions. This is only safe if you then retrain it based on any
446 errors you discover.
447 SpamAssassin does not support this method, due to experimental
449 results which strongly indicate that it does not work well, and
450 since Bayes is only one part of the resulting score presented to
451 the user (while Bayes may have made the wrong decision about a
452 mail, it may have been overridden by another system).
453 3. Unsupervised learning from SpamAssassin rules
455 Also called 'auto-learning' in SpamAssassin. Based on statistical
456 analysis of the SpamAssassin success rates, we can automatically
457 train the Bayesian database with a certain degree of confidence
458 that our training data is accurate.
459 It should be supplemented with some supervised training in
461 addition, if possible.
462 This is the default, but can be turned off by setting the
464 SpamAssassin configuration parameter "bayes_auto_learn" to 0.
465 4. Mistake-based training
467 This means training on a small number of mails, then only training
468 on messages that SpamAssassin classifies incorrectly. This works,
469 but it takes longer to get it right than a full training session
470 would.
471
473 sa-learn and the other parts of SpamAssassin's Bayesian learner, use a
474 set of persistent database files to store the learnt tokens, as
475 follows.
476 bayes_toks
478 The database of tokens, containing the tokens learnt, their count
479 of occurrences in ham and spam, and the timestamp when the token
480 was last seen in a message.
481 This database also contains some 'magic' tokens, as follows: the
483 version number of the database, the number of ham and spam messages
484 learnt, the number of tokens in the database, and timestamps of:
485 the last journal sync, the last expiry run, the last expiry token
486 reduction count, the last expiry timestamp delta, the oldest token
487 timestamp in the database, and the newest token timestamp in the
488 database.
489 This is a database file, using "DB_File". The database 'version
491 number' is 0 for databases from 2.5x, 1 for databases from certain
492 2.6x development releases, 2 for 2.6x, and 3 for 3.0 and later
493 releases.
494 bayes_seen
496 A map of Message-Id and some data from headers and body to what
497 that message was learnt as. This is used so that SpamAssassin can
498 avoid re-learning a message it has already seen, and so it can
499 reverse the training if you later decide that message was learnt
500 incorrectly.
501 This is a database file, using "DB_File".
503 bayes_journal
505 While SpamAssassin is scanning mails, it needs to track which
506 tokens it uses in its calculations. To avoid the contention of
507 having each SpamAssassin process attempting to gain write access to
508 the Bayes DB, the token timestamps are written to a 'journal' file
509 which will later (either automatically or via "sa-learn --sync") be
510 used to synchronize the Bayes DB.
511 Also, through the use of "bayes_learn_to_journal", or when using
513 the "--no-sync" option with sa-learn, the actual learning data will
514 take be placed into the journal for later synchronization. This is
515 typically useful for high-traffic sites to avoid the same
516 contention as stated above.
517
519 Since SpamAssassin can auto-learn messages, the Bayes database files
520 could increase perpetually until they fill your disk. To control this,
521 SpamAssassin performs journal synchronization and bayes expiration
522 periodically when certain criteria (listed below) are met.
523 SpamAssassin can sync the journal and expire the DB tokens either
525 manually or opportunistically. A journal sync is due if --sync is
526 passed to sa-learn (manual), or if the following is true
527 (opportunistic):
528 - bayes_journal_max_size does not equal 0 (means don't sync)
530 - the journal file exists
531 and either:
533 - the journal file has a size greater than bayes_journal_max_size
535 or
537 - a journal sync has previously occurred, and at least 1 day has passed
539 since that sync
540 Expiry is due if --force-expire is passed to sa-learn (manual), or if
542 all of the following are true (opportunistic):
543 - the last expire was attempted at least 12hrs ago
545 - bayes_auto_expire does not equal 0
546 - the number of tokens in the DB is > 100,000
547 - the number of tokens in the DB is > bayes_expiry_max_db_size
548 - there is at least a 12 hr difference between the oldest and newest
549 token atimes
550 EXPIRE LOGIC
552 If either the manual or opportunistic method causes an expire run to
553 start, here is the logic that is used:
554 - figure out how many tokens to keep. take the larger of either
556 bayes_expiry_max_db_size * 75% or 100,000 tokens. therefore, the goal
557 reduction is number of tokens - number of tokens to keep.
558 - if the reduction number is < 1000 tokens, abort (not worth the
559 effort).
560 - if an expire has been done before, guesstimate the new atime delta
561 based on the old atime delta. (new_atime_delta = old_atime_delta *
562 old_reduction_count / goal)
563 - if no expire has been done before, or the last expire looks "weird",
564 do an estimation pass. The definition of "weird" is:
565 - last expire over 30 days ago
566 - last atime delta was < 12 hrs
567 - last reduction count was < 1000 tokens
568 - estimated new atime delta is < 12 hrs
569 - the difference between the last reduction count and the goal
570 reduction count is > 50%
571 ESTIMATION PASS LOGIC
573 Go through each of the DB's tokens. Starting at 12hrs, calculate
574 whether or not the token would be expired (based on the difference
575 between the token's atime and the db's newest token atime) and keep the
576 count. Work out from 12hrs exponentially by powers of 2. ie: 12hrs *
577 1, 12hrs * 2, 12hrs * 4, 12hrs * 8, and so on, up to 12hrs * 512
578 (6144hrs, or 256 days).
579 The larger the delta, the smaller the number of tokens that will be
581 expired. Conversely, the number of tokens goes up as the delta gets
582 smaller. So starting at the largest atime delta, figure out which
583 delta will expire the most tokens without going above the goal
584 expiration count. Use this to choose the atime delta to use, unless
585 one of the following occurs:
586 - the largest atime (smallest reduction count) would expire too many
588 tokens. this means the learned tokens are mostly old and there needs
589 to be new tokens learned before an expire can occur.
590 - all of the atime choices result in 0 tokens being removed. this means
591 the tokens are all newer than 12 hours and there needs to be new tokens
592 learned before an expire can occur.
593 - the number of tokens that would be removed is < 1000. the benefit
594 isn't worth the effort. more tokens need to be learned.
595 If the expire run gets past this point, it will continue to the end. A
597 new DB is created since the majority of DB libraries don't shrink the
598 DB file when tokens are removed. So we do the "create new, migrate old
599 to new, remove old, rename new" shuffle.
600 EXPIRY RELATED CONFIGURATION SETTINGS
602 "bayes_auto_expire" is used to specify whether or not SpamAssassin
603 ought to opportunistically attempt to expire the Bayes database. The
604 default is 1 (yes).
605 "bayes_expiry_max_db_size" specifies both the auto-expire token count
606 point, as well as the resulting number of tokens after expiry as
607 described above. The default value is 150,000, which is roughly
608 equivalent to a 6Mb database file if you're using DB_File.
609 "bayes_journal_max_size" specifies how large the Bayes journal will
610 grow before it is opportunistically synced. The default value is
611 102400.
612
614 The sa-learn command is part of the Mail::SpamAssassin Perl module.
615 Install this as a normal Perl module, using "perl -MCPAN -e shell", or
616 by hand.
617
619 spamassassin(1) spamc(1) Mail::SpamAssassin(3)
620 Mail::SpamAssassin::ArchiveIterator(3)
621 <http://www.paulgraham.com/> Paul Graham's "A Plan For Spam" paper
623 <http://radio.weblogs.com/0101454/stories/2002/09/16/spamDetection.html>
625 Gary Robinson's f(x) and combining algorithms, as used in SpamAssassin
626 <http://www.bgl.nu/~glouis/bogofilter/> 'Training on error' page. A
628 discussion of various Bayes training regimes, including 'train on
629 error' and unsupervised training.
630
632 "Mail::SpamAssassin"
633
635 The SpamAssassin(tm) Project <http://spamassassin.apache.org/>
636perl v5.10.1 2013-12-05 SA-LEARN(1)