SS Recently Letting MUCH More Through

I have been a SS user for many years and feel it is a wonderful program, the best of its class.

I have made no changes in my email program (GyazMail) or OS (10.4.x, now 11) in recent months and I have the most up-to-date SS. In the last few weeks, SS has performed much moor poorly than it ever has.

Where I used to get maybe 2-3 spam messages invading my inbox a month, i now get 5-15 a day! The Statistics tell the tale: from when I started using SS to now, the “SpamSieve Accuracy” registers 97.9% correct; in the last two weeks, that stat deteriorated to 85.3%.

I’m puzzled why the degradation occurred, and, even more urgent, want my SS with peak performance back.

Just now, I have set the spam-catching strategy preference to the “most aggressive” setting. Is there anything else I can do? Why is this happening since the beginning of the year?

Please send me the log file and the false negatives files so that I can see what’s happening.

According to the log file that you sent me, SpamSieve’s corpus contains only 6 good messages and 8 spam ones. So it has very little information to go on. Yesterday it had 127 spam messages and 82 good ones (also not very many), but when you launched it again today the corpus was empty again. It looks like this same pattern of the corpus being erased between launches repeats going back several weeks. My guess is that there is a permissions problem on the folder:

/Users/<username>/Library/Application Support/SpamSieve/

or on the Corpus.corpus package itself that’s preventing SpamSieve from saving to (or perhaps reading from) disk.

Can I resolve that by running Repair Permissions (which I will try now)? And would it help to reinstall SS?

Thanks.

Probably not. I think that only affects directories that are owned by the system. I’d start by using Get Info on the SpamSieve folder in Application Support and on the Corpus.corpus file and making sure that those both have read-write permission.

I’m not sure that it’s a permissions problem, though, since normally SpamSieve would report an error in that case, and I didn’t see any in the log. I’d definitely run a disk repair program in case your disk’s catalog is damaged and it’s losing files.

No.

Everything is Read and Write; Disk Utility shows the disk is okay.

I’ll see how things go for the next week or two and report back.

Thanks for your help.

Please monitor the corpus file to see whether the messages SpamSieve is trained with are saved when it quits. And I suggest making backups of it, too.

Buried in Spam again too
I’m experiencing something very similar to Che-Dog, but with some notable differences. LIke CD, I have been using SpamSieve for many years. I have spent time training the product, and it USED to work very effectively.

This morning, a Monday, SpamSieve correctly identified about 25 messages as spam… and let through about 50 messages, many of which are VERY OBVIOUSLY spam given that they have undisguised text strings like “men’s health products” and “genuine viagra”, correctly spelled, in the body and the subject. This is typical of the volume and pattern I experience later, and the percentage of correctly-identified spam messages is consistently below 50%.

Based on this thread, I peeked at my corpus, which has over 5,000 examples, many of them tagged with today’s date as “last used.” I don’t think my permissions are screwed up the way it appeared Che-Dog’s was.

I came to the site today in utter frustration. Should I clear my corpus and start over from scratch? Is there something else I can do? I’m growing weary enough of sorting through all this trash that I give passing thoughts to actually changing my eMail address, but from a business perspective this would be quite unwise for me. Please help. I am attaching my log to this post; at least I am attempting to; it’s been 5 minutes and it is still uploading my 2.7 MB log file. Yeah, that didn’t work, so I’ll eMail it per the instructions on this site.

The log contains personal information such as the senders and subjects of the messages that you’ve received, so you probably don’t want to post it, anyway.

Looking at your log, you do not seem to be experiencing the same problem as Che-Dog; your corpus contains almost 6,000 messages.

First, it looks like you’ve set SpamSieve to be more Conservative about classifying messages as spam. I suggest that you move the Spam-catching strategy slider back to the middle (or maybe a little towards Aggressive).

Second, about half of the messages that you trained as spam were not ones that SpamSieve had thought were good. That is, the log shows a “Trained: Spam (Manual)” entry but no corresponding “Predicted: Good” entry. This usually indicates a setup problem in your mail program.

According to the log, SpamSieve let 10 spam messages through today, and you trained 22 as spam. So it would seem that either your guess of 50 is overstated or that you haven’t been telling SpamSieve about all the spam messages that got through.

I suggest that you fix the slider and the PowerMail setup first. Then, only if necessary, would I recommend resetting the corpus and re-training SpamSieve.

Thank you for the very rapid reply, Michael; also thank you for thoroughly addressing these situations rather than offering boilerplate responses as so many other companies do.

I have addressed your points, above: in fact, I haven’t been bothering to mark every message as spam; many I just throw away; the ones I mark are the ones it seems the program really should be getting.

Upon reading your response, I located an obvious spam message in PowerMail, selected it and ran the Spam: evaluate filter on it. NO spam rating whatsoever. I then went in and adjusted the slider up one point, and edited the filter so that it is only looking to see if the recipient is in my address book first. Having done this – and it was allowing stuff past the filter if I had a “previous recipient” that matched, uh-oh – I repeated the filter test on the same message, and it was immediately rated a 96.

So yeah; I think you nailed it. I’ll come back here and report results later, either way, after some time has gone by. Further analysis of the log suggests that there are some errant entries in the “Spamsieve Whitelist”, so I will attempt to find that and edit it as well.

; ) David

I cannot overstate how strongly I recommend that you correct every mistake that SpamSieve makes. Do not just delete the messages. If a spam message gets put in the inbox and you don’t tell SpamSieve that it’s spam, this is almost like training SpamSieve that the message is good and adding whitelist rule(s) for it. (In fact, this probably explains the errant entries that you noticed.)

Good.

Better!!
OMG, it is working SO much better now having addressed the above per your instructions. Thank you. It’s like having a whole new spam filter.

; ) DF