Virus Bulletin
Copyright © 2023 Virus Bulletin
In the Q4 2023 VBSpam test – which forms part of Virus Bulletin’s continuously running security product test suite – we measured the performance of a number of email security solutions against various streams of wanted, unwanted and malicious emails. One third of the solutions we tested opted to be included in the public test, the rest opting for private testing (all details and results remaining unpublished). The solutions tested publicly were eight full email security solutions, one custom configured solution1, one open-source solution and one blocklist.
Out of the eight full email security solutions, six blocked 100% of malware samples while the other two blocked more than 99.50%. It seems that emails containing malicious attachments no longer present much of a problem. Microsoft’s decision to disable macros by default in Office apps for Windows users may have contributed to bad actors’ lack of interest in this kind of threat.
However, though not on quite the same level, we also see a good performance from the security solutions against phishing emails, with detection rates for the most part higher than 99.50%.
For some additional background to this report, the table and map below show the geographical distribution (based on sender IP address) of the spam emails seen in the test. (Note: these statistics are relevant only to the spam samples we received during the test period.)
# | Sender's IP country | Percentage of spam |
1 | Kazakhstan | 20.69% |
2 | Russian Federation | 8.93% |
3 | Uzbekistan | 6.36% |
4 | China | 5.98% |
5 | Kyrgyzstan | 3.93% |
6 | Pakistan | 3.68% |
7 | United States | 3.40% |
8 | Islamic Republic of Iran | 2.64% |
9 | Vietnam | 2.64% |
10 | India | 2.45% |
We continue to see the same trend as we noted in previous reports: most of the missed phishing samples are in languages other than English. The following screenshots show the phishing emails that evaded the filters of the majority of the tested solutions: Norwegian, Romanian, German and Hungarian samples.
From time to time, we see short duration spam campaigns that are missed by the majority of the solutions we test. The emails advertise certain stocks that historically reached high values but which have now dropped to less than $1. There are no URLs or attachments, and many of the words seem to be skewed on purpose, to make it challenging to filter.
Most of the tested solutions managed to successfully block all the malware samples. There was only one challenging sample that warrants a mention here.
The attachment contains a ‘.doc’ file using a macro to access an external link that downloads the payload, a ‘.exe’ file. The ‘.exe’ file is reported2 to be linked to the LokiBot malware.
The majority of the tested solutions managed to achieve high catch rates both on overall spam samples and on the malware sub-category, with values higher than 99%. In particular, we highlight the performance of SEPPmail.cloud Filter, which missed only one phishing sample.
Of the participating full solutions, two achieved a VBSpam award – SEPPmail.cloud Filter and Zoho Mail – while six – Bitdefender GravityZone Premium, FortiMail, Mimecast, N-able Mail Assure, N-Able SpamExperts and Net At Work NoSpamProxy – were awarded a VBSpam+ certification, as was the custom configured solution Spamhaus DQS.
(Note: since, for a number of products, catch rates and/or final scores were very close to, whilst remaining a fraction below, 100%, in this test we quote all the spam-related scores with three decimal places.)
SC rate: 99.990%
|
It was another impressive performance from Bitdefender, and the vendor’s uninterrupted series of VBSpam+ awards continues in the Q4 2023 VBSpam test. While blocking 99.99% of the spam samples, the product correctly filtered all of the legitimate samples.
SC rate: 99.995%
|
It was an almost perfect performance from Fortinet in this test: no false positives, a 100% catch rate on both malware and phishing, and only 13 false negatives. The product easily earns its VBSpam+ certification.
SC rate: 99.967%
|
No malware sample was able to get past Mimecast’s filters, and there were no false positives of any kind. With a final score of 99.967, Mimecast earns VBSpam+ certification for its performance in the Q4 2023 VBSpam Test.
SC rate: 99.982%
|
N-able Mail Assure continued the run of good performances in this test, earning the product another VBSpam+ award. Beside the 100% malware catch rate we highlight the lack of ham false positives and the higher than 99% phishing catch rate.
SC rate: 99.982%
|
With identical scores to its sister product, N-able SpamExperts also earns VBSpam+ certification in this test.
SC rate: 99.988%
|
It was another balanced performance for Net at Work’s email security solution, resulting in another VBSpam+ award to add to its collection. Particularly of note were the lack of false positives and the higher than 99.9% spam catch rate.
SC rate: 98.750%
|
The highlight for Rspamd in the Q4 2023 VBSpam Test was its 98.75% spam catch rate – the highest value achieved by the open-source product this year and one of the best since it joined the VBSpam test.
SC rate: 99.998%
|
Only four spam samples were missed by SEPPmail.cloud Filter – a truly impressive performance and the best across all the 2023 quarterly VBSpam tests. A number of false positives brought the final score down a touch, but the product easily earns VBSpam certification.
SC rate: 99.902%
|
Spamhaus SpamAssassin Data Query Service (DQS) is a custom configured solution that integrates the Spamhaus DQS DNSBL service and the free open-source solution SpamAssassin. In this test no ham or newsletter samples were blocked by this combined solution. With a final score of 99.902 the product earns VBSpam+ certification.
SC rate: 99.816%
|
Zoho Mail achieved higher than 99% catch rates not only on the overall spam samples but also on the malware and phishing sub-categories. Despite some false positives the product reached a final score of 99.557 and earns a VBSpam award.
SC rate: 99.144% |
Abusix Mail Intelligence is a set of blocklists that is tested as a partial solution because it has access only to parts of the emails (IP addresses, domains, URLs), which are queried to their DNS zones. With this setup, the solution’s 99.144% spam catch rate and lack of ham false positives are commendable.
True negatives | False positives | FP rate | False negatives | True positives | SC rate | Final score | VBSpam | |
Bitdefender GravityZone Premium | 2297 | 0 | 0.00% | 27 | 261847 | 99.990% | 99.990 | |
Fortinet FortiMail | 2297 | 0 | 0.00% | 13 | 261861 | 99.995% | 99.995 | |
Mimecast | 2297 | 0 | 0.00% | 87.6 | 261786.4 | 99.967% | 99.967 | |
N-able Mail Assure | 2297 | 0 | 0.00% | 46.4 | 261827.6 | 99.982% | 99.982 | |
N-able SpamExperts | 2297 | 0 | 0.00% | 46.4 | 261827.6 | 99.982% | 99.982 | |
Net At Work NoSpamProxy | 2297 | 0 | 0.00% | 31 | 261843 | 99.988% | 99.945 | |
Rspamd | 2274 | 23 | 1.00% | 3272.8 | 258601.2 | 98.750% | 93.688 | |
SEPPmail.cloud Filter | 2296 | 1 | 0.04% | 4.2 | 261869.8 | 99.998% | 99.739 | |
Spamhaus DQS + SpamAssassin‡ | 2297 | 0 | 0.00% | 257.2 | 261616.8 | 99.902% | 99.902 | |
Zoho Mail | 2296 | 1 | 0.04% | 480.8 | 261393.2 | 99.816% | 99.557 | |
Abusix Mail Intelligence* | 2297 | 0 | 0.00% | 2240.8 | 259633.2 | 99.144% | 99.144 | N/A |
‡Spamhaus Data Query Service (DQS) + SpamAssassin is a fully configured solution that integrates Spamhaus DQS on top of SpamAssassin. Spamhaus DQS is not a stand-alone solution but rather a DNSBL service that can be added to MTAs and email security solutions such as SpamAssasssin. The test set up reflects the real-life performance expected from this combined production deployment, not as individual product elements.
*This product is a partial solution and its performance should not be compared with that of other products.
Newsletters | Malware | Phishing | Project Honey Pot | Abusix | MXMailData | STDev† | |||||||
False positives | FP rate | False negatives | SC rate | False negatives | SC rate | False negatives | SC rate | False negatives | SC rate | False negatives | SC rate | ||
Bitdefender GravityZone Premium | 0 | 0.0% | 5 | 99.800% | 6 | 99.770% | 2 | 99.985% | 20 | 99.992% | 5 | 99.770% | 0.08 |
Fortinet FortiMail | 0 | 0.0% | 0 | 100.000% | 0 | 100.000% | 9 | 99.931% | 4 | 99.998% | 0 | 100.000% | 0.17 |
Mimecast | 0 | 0.0% | 0 | 100.000% | 6 | 99.770% | 62.6 | 99.519% | 25 | 99.990% | 0 | 100.000% | 0.52 |
N-able Mail Assure | 0 | 0.0% | 0 | 100.000% | 13 | 99.510% | 22 | 99.831% | 24.4 | 99.990% | 0 | 100.000% | 0.2 |
N-able SpamExperts | 0 | 0.0% | 0 | 100.000% | 13 | 99.510% | 22 | 99.831% | 24.4 | 99.990% | 0 | 100.000% | 0.2 |
Net At Work NoSpamProxy | 1 | 1.4% | 0 | 100.000% | 6 | 99.770% | 6 | 99.954% | 25 | 99.990% | 0 | 100.000% | 0.21 |
Rspamd | 2 | 2.8% | 877 | 65.740% | 282 | 89.410% | 1356.2 | 89.585% | 1092.6 | 99.557% | 824 | 61.600% | 4.36 |
SEPPmail.cloud Filter | 1 | 1.4% | 0 | 100.000% | 1 | 99.960% | 4.2 | 99.968% | 0 | 100.000% | 0 | 100.000% | 0.2 |
Spamhaus DQS + SpamAssassin‡ | 0 | 0.0% | 82 | 96.800% | 55 | 97.930% | 8.2 | 99.937% | 164 | 99.934% | 85 | 96.040% | 0.68 |
Zoho Mail | 1 | 1.4% | 8 | 99.690% | 7 | 99.740% | 204.8 | 98.427% | 268 | 99.891% | 8 | 99.630% | 1.02 |
Abusix Mail Intelligence* | 0 | 0.0% | 703 | 72.540% | 59 | 97.780% | 1274.8 | 90.210% | 318 | 99.871% | 648 | 69.800% | 3.91 |
†The standard deviation of a product is calculated using the set of its hourly spam catch rates.
‡Spamhaus Data Query Service (DQS) + SpamAssassin is a fully configured solution that integrates Spamhaus DQS on top of SpamAssassin. Spamhaus DQS is not a stand-alone solution but rather a DNSBL service that can be added to MTAs and email security solutions such as SpamAssasssin. The test set up reflects the real-life performance expected from this combined production deployment, not as individual product elements.
*This product is a partial solution and its performance should not be compared with that of other products. None of the queries to the IP blocklist included any information on the attachments; hence its performance on the malware corpus is added purely for information.
Speed | ||||
10% | 50% | 95% | 98% | |
Bitdefender GravityZone Premium | ||||
Fortinet FortiMail | ||||
Mimecast | ||||
N-able Mail Assure | ||||
N-able SpamExperts | ||||
Net At Work NoSpamProxy | ||||
Rspamd | ||||
SEPPmail.cloud Filter | ||||
Spamhaus Data Query Service (DQS) + SpamAssassin‡ | ||||
Zoho Mail |
0-30 seconds | 30 seconds to two minutes | two minutes to 10 minutes | more than 10 minutes |
‡Spamhaus Data Query Service (DQS) + SpamAssassin is a fully configured solution that integrates Spamhaus DQS on top of SpamAssassin. Spamhaus DQS is not a stand-alone solution but rather a DNSBL service that can be added to MTAs and email security solutions such as SpamAssasssin. The test set up reflects the real-life performance expected from this combined production deployment, not as individual product elements.
Products ranked by final score | |
Fortinet FortiMail | 99.995 |
Bitdefender GravityZone Premium | 99.990 |
N-able Mail Assure | 99.982 |
N-able SpamExperts | 99.982 |
Mimecast | 99.967 |
Net At Work NoSpamProxy | 99.945 |
Spamhaus DQS + SpamAssassin‡ | 99.902 |
SEPPmail.cloud Filter | 99.739 |
Zoho Mail | 99.557 |
Rspamd | 93.688 |
‡Spamhaus Data Query Service (DQS) + SpamAssassin is a fully configured solution that integrates Spamhaus DQS on top of SpamAssassin. Spamhaus DQS is not a stand-alone solution but rather a DNSBL service that can be added to MTAs and email security solutions such as SpamAssasssin. The test set up reflects the real-life performance expected from this combined production deployment, not as individual product elements.
Hosted solutions | Anti-malware | IPv6 | DKIM | SPF | DMARC | Multiple MX-records | Multiple locations |
Mimecast | Mimecast | √ | √ | √ | √ | √ | |
N-able Mail Assure | N-able Mail Assure | √ | √ | √ | √ | ||
N-able SpamExperts | SpamExperts | √ | √ | √ | √ | ||
Net At Work NoSpamProxy | 32Guards & NoSpamProxy | √ | √ | √ | √ | √ | |
SEPPmail.cloud Filter | SEPPmail | √ | √ | √ | √ | √ | √ |
Zoho Mail | Zoho | √ | √ | √ | √ | √ |
Local solutions | Anti-malware | IPv6 | DKIM | SPF | DMARC | Interface | |||
CLI | GUI | Web GUI | API | ||||||
Bitdefender GravityZone Premium | Bitdefender | √ | √ | √ | √ | ||||
Fortinet FortiMail | Fortinet | √ | √ | √ | √ | √ | √ | √ | |
Rspamd | None | √ | |||||||
Spamhaus Data Query Service (DQS) + SpamAssassin‡ | Optional | √ | √ | √ | √ |
‡Spamhaus Data Query Service (DQS) + SpamAssassin is a fully configured solution that integrates Spamhaus DQS on top of SpamAssassin. Spamhaus DQS is not a stand-alone solution but rather a DNSBL service that can be added to MTAs and email security solutions such as SpamAssasssin. The test set up reflects the real-life performance expected from this combined production deployment, not as individual product elements.
The full VBSpam test methodology can be found at https://46.101.67.4/testing/vbspam/vbspam-methodology/vbspam-methodology-ver20.
The test ran for 16 days, from 12am on 4 November to 12am on 20 November 2023 (GMT).
The test corpus consisted of 264,270 emails. 261,902 of these were spam, 13,039 of which were provided by Project Honey Pot, 246,717 were provided by Abusix with the remaining 2,146 spam emails provided by MXMailData. There were 2,297 legitimate emails (‘ham’) and 71 newsletters, a category that includes various kinds of commercial and non-commercial opt-in mailings.
35 emails in the spam corpus were considered ‘unwanted’ (see the June 2018 report) and were included with a weight of 0.2; this explains the non-integer numbers in some of the tables.
Moreover, 2,560 emails from the spam corpus were found to contain a malicious attachment while 2,662 contained a link to a phishing or malware site; though we report separate performance metrics on these corpora, it should be noted that these emails were also counted as part of the spam corpus.
Emails were sent to the products in real time and in parallel. Though products received the email from a fixed IP address, all products had been set up to read the original sender’s IP address as well as the EHLO/HELO domain sent during the SMTP transaction, either from the email headers or through an optional XCLIENT SMTP command3.
For those products running in our lab, we all ran them as virtual machines on a VMware ESXi cluster. As different products have different hardware requirements – not to mention those running on their own hardware, or those running in the cloud – there is little point comparing the memory, processing power or hardware the products were provided with; we followed the developers’ requirements and note that the amount of email we receive is representative of that received by a small organization.
Although we stress that different customers have different needs and priorities, and thus different preferences when it comes to the ideal ratio of false positive to false negatives, we created a one-dimensional ‘final score’ to compare products. This is defined as the spam catch (SC) rate minus five times the weighted false positive (WFP) rate. The WFP rate is defined as the false positive rate of the ham and newsletter corpora taken together, with emails from the latter corpus having a weight of 0.2:
WFP rate = (#false positives + 0.2 * min(#newsletter false positives , 0.2 * #newsletters)) / (#ham + 0.2 * #newsletters)
while in the spam catch rate (SC), emails considered ‘unwanted’ (see above) are included with a weight of 0.2.
The final score is then defined as:
Final score = SC - (5 x WFP)
In addition, for each product, we measure how long it takes to deliver emails from the ham corpus (excluding false positives) and, after ordering these emails by this time, we colour-code the emails at the 10th, 50th, 95th and 98th percentiles:
(green) = up to 30 seconds | |
(yellow) = 30 seconds to two minutes | |
(orange) = two to ten minutes | |
(red) = more than ten minutes |
Products earn VBSpam certification if the value of the final score is at least 98 and the ‘delivery speed colours’ at 10 and 50 per cent are green or yellow and that at 95 per cent is green, yellow or orange.
Meanwhile, products that combine a spam catch rate of 99.5% or higher with a lack of false positives, no more than 2.5% false positives among the newsletters and ‘delivery speed colours’ of green at 10 and 50 per cent and green or yellow at 95 and 98 per cent earn a VBSpam+ award.
Meanwhile, products that combine a spam catch rate of 99.5% or higher with a lack of false positives, no more than 2.5% false positives among the newsletters and ‘delivery speed colours’ of green at 10 and 50 per cent and green or yellow at 95 and 98 per cent earn a VBSpam+ award.
1 Spamhaus Data Query Service (DQS) + SpamAssassin is a custom solution built on top of the SpamAssassin open-source anti-spam platform.
2 https://www.virustotal.com/gui/file/30de8327003cc6cb8d30fdea5758aafc01481714fd9c280278bafc70bb36f2cc/behavior.