« Return to: Natural-Born Spam Killers
| Integrated Spam Filter Utility | List price | Before training: Spam caught | Before training: False positives1 | After training: Spam caught | After training: False positives1 | Comments |
Cloudmark SpamNet | Free beta2 | 66% | 5% | 65%3 | 8% | Uses a peer-to-peer network to filter. Scanning multiple messages is slow. An Outlook Express plug-in will be available soon. |
| Best Buy IHateSpam for Outlook 3.2  | $20 | 94% | 30% | 98% | 18% | Inexpensive and easy-to-use package offers accurate filtering. Only product backed by toll-free phone support. |
IHateSpam for Outlook Express 3.2 | $20 | 95% | 28% | 94% | 45% | Outlook Express version of our Best Buy doesn't work as smoothly and produced many false positives. |
| Spamnix for Eudora | $30 | 66% | 5% | 99% | 13% | Outstanding performance after tweaking, but creating filters can prove time-consuming. An Outlook version will be available soon. |
| Stand-alone Spam Filter Utility | List price | Before training: Spam caught | Before training: False positives1 | After training: Spam caught | After training: False positives1 | Comments |
Firetrust MailWasher Pro 3 | $30 | 7%4 | 0% | 96% | 10% | Performed well after training, though it crashed occasionally. Supports POP3 and MSN Hotmail. |
| Lyris MailShield Desktop 3 | $60 | 78% | 8% | 98% | 22% | Points-based filtering has trouble with legitimate HTML e-mail. Supports POP3, MAPI, IMAP, MSN Hotmail, and some Yahoo e-mail. |
McAfee SpamKiller 4 | $40 download, $50 CD-ROM | 84% | 23% | 98% | 2% | Setup and training are particularly simple and intuitive. Supports any POP3 or IMAP client and MSN Hotmail. |
How We Test: To test each utility, we restored two backups of a real mail file containing both spam and legitimate e-mail. After installing a tool, we restored the first backup, which had 683 messages, including 623 we identified as spam. We recorded the amount of spam the filter caught and the number of legitimate messages it wrongly flagged at its default settings. We trained each program on this first set of mail by creating filters to block missed spam and allow incorrectly flagged mail. Where applicable, we switched to a stricter filter setting. We then restored the second backup--which included the first batch plus an additional 269 messages--and tested the filter with the new settings.
Chart Note: Bold percentages indicate the best performance.
1 Legitimate e-mail wrongly flagged as spam.
2 Paid final release coming soon.
3 It may take a week or more for training to affect SpamNet's performance.
4 199 spam messages (about 32 percent) marked as possible.
To test each utility, we restored two backups of a real mail file containing both spam and legitimate e-mail. After installing a tool, we restored the first backup, which had 683 messages, including 623 we identified as spam. We recorded the amount of spam the filter caught and the number of legitimate messages it wrongly flagged at its default settings. We trained each program on this first set of mail by creating filters to block missed spam and allow incorrectly flagged mail. Where applicable, we switched to a stricter filter setting. We then restored the second backup--which included the first batch plus an additional 269 messages--and tested the filter with the new settings. Chart Note: Bold percentages indicate the best performance.