VBSpam Email Security Comparative Review March 2021

Ionuţ Răileanu

Virus Bulletin

Copyright © 2021 Virus Bulletin

Introduction

In this test, which forms part of Virus Bulletin’s continuously running security product test suite, 28 email security products were assembled on the test bench to measure their performance against various streams of wanted, unwanted and malicious emails. The results for the 10 products that elected to be tested publicly – seven full email security solutions, one custom configured solution¹, one open-source solution and one blocklist – are included in this report.

In this round of testing we welcome three new participants to the public VBSpam test: Cleanmail Domain Gateway, Cyren eXpurgate and Rspamd.

The results are good – the security solutions we tested demonstrated that they are well equipped to defend in the first line against malware and phishing attacks.

With the threat landscape shifting due to the takedown of the Emotet botnet we noted an overall decrease in the number of spam emails seen, not only of those with malicious attachments or URLs. In this report we briefly describe the emails that proved the most challenging for the tested solutions to deal with correctly.

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 in real time.)

Top 10 countries from which spam was sent.

Geographical distribution of spam based on sender IP address.

Malware and phishing

QBot malware

This malware campaign was the most challenging for the solutions participating in our test. The attached zip archive contained an xls file with a macro. On opening the file, a connection was initiated to a domain associated with QBot (casadodestino[.]com). The given example had an attachment with the SHA 256 value of 1da459e6c1bb2779ec85ecab38dc22bcbba40376af754f4ff9ac93bb5b532a22.

Three solutions managed to block every email of this campaign: Cleanmail Domain Gateway, Libraesva and ZEROSPAM.

Legitimate services in phishing emails

Despite a decrease in overall numbers, the phishing emails that are the most successful in evading email security solution filtering continue to be those that contain legitimate service URLs, an example of which is shown below. At the time of the analysis the URL in this sample was blocked by Google for violating its Terms of Use so we don’t have any more details on its behaviour.

The solutions that correctly blocked this challenging sample were Bitdefender, Cleanmail Domain Gateway and ZEROSPAM.

Results

The majority of the tested security solutions managed to block more than 99% of the spam emails, with three products, Cleanmail, Libraesva and ZEROSPAM, blocking all the malware samples. In the phishing category, the catch rates were lower but we still see many values of more than 98%.

Of the participating full solutions, four achieved a VBSpam award: Axway, Cleanmail, Fortinet and ZEROSPAM, while a further four performed well enough to achieve a VBSpam+ award: Bitdefender, Cyren eXpurgate, Libraesva and Spamhaus DQS.

Axway MailGate 5.6

SC rate: 99.86%
FP rate: 0.03%
Final score: 99.73
Malware catch rate: 96.41%
Phishing catch rate: 98.58%
Project Honey Pot SC rate: 99.72%
Abusix SC rate: 99.88%
Newsletters FP rate: 0.0%
Speed:

10%

50%

95%

98%

Bitdefender Security for Mail Servers 3.1.7

SC rate: 99.93%
FP rate: 0.00%
Final score: 99.93
Malware catch rate: 97.26%
Phishing catch rate: 99.07%
Project Honey Pot SC rate: 99.99%
Abusix SC rate: 99.92%
Newsletters FP rate: 0.0%
Speed:

10%

50%

95%

98%

Cleanmail Domain Gateway

SC rate: 99.98%
FP rate: 0.03%
Final score: 99.75
Malware catch rate: 100.00%
Phishing catch rate: 99.95%
Project Honey Pot SC rate: 99.97%
Abusix SC rate: 99.98%
Newsletters FP rate: 3.3%
Speed:

10%

50%

95%

98%

Cyren eXpurgate

SC rate: 99.65%
FP rate: 0.00%
Final score: 99.62
Malware catch rate: 98.68%
Phishing catch rate: 94.77%
Project Honey Pot SC rate: 98.93%
Abusix SC rate: 99.78%
Newsletters FP rate: 0.8%
Speed:

10%

50%

95%

98%

Fortinet FortiMail

SC rate: 99.78%
FP rate: 0.03%
Final score: 99.65
Malware catch rate: 99.06%
Phishing catch rate: 98.64%
Project Honey Pot SC rate: 99.47%
Abusix SC rate: 99.84%
Newsletters FP rate: 0.0%
Speed:

10%

50%

95%

98%

Libraesva ESG v.4.7

SC rate: 99.92%
FP rate: 0.00%
Final score: 99.90
Malware catch rate: 100.00%
Phishing catch rate: 99.67%
Project Honey Pot SC rate: 99.74%
Abusix SC rate: 99.96%
Newsletters FP rate: 0.8%
Speed:

10%

50%

95%

98%

Rspamd

SC rate: 90.17%
FP rate: 0.38%
Final score: 88.08
Malware catch rate: 65.63%
Phishing catch rate: 75.15%
Project Honey Pot SC rate: 84.51%
Abusix SC rate: 91.24%
Newsletters FP rate: 5.7%
Speed:

10%

50%

95%

98%

Spamhaus Data Query Service

SC rate: 99.71%
FP rate: 0.00%
Final score: 99.64
Malware catch rate: 93.48%
Phishing catch rate: 98.53%
Project Honey Pot SC rate: 99.17%
Abusix SC rate: 99.82%
Newsletters FP rate: 2.5%
Speed:

10%

50%

95%

98%

ZEROSPAM

SC rate: 98.75%
FP rate: 0.00%
Final score: 98.73
Malware catch rate: 100.00%
Phishing catch rate: 98.86%
Project Honey Pot SC rate: 98.62%
Abusix SC rate: 98.78%
Newsletters FP rate: 0.8%
Speed:

10%

50%

95%

98%

Abusix Mail Intelligence²

SC rate: 98.99%
FP rate: 0.03%
Final score: 98.84
Malware catch rate: 98.02%
Phishing catch rate: 98.20%
Project Honey Pot SC rate: 97.42%
Abusix SC rate: 99.29%
Newsletters FP rate: 0.8%

Results tables

	True negatives	False positives	FP rate	False negatives	True positives	SC rate	Final score	VBSpam
Axway	3908	1	0.03%	281	194469.4	99.86%	99.73
Bitdefender	3909	0	0.00%	127.8	194622.6	99.93%	99.93
Cleanmail Domain Gateway	3908	1	0.03%	43.8	194706.6	99.98%	99.75
Cyren eXpurgate	3909	0	0.00%	689.4	194061	99.65%	99.62
FortiMail	3908	1	0.03%	433.2	194317.2	99.78%	99.65
Libraesva	3909	0	0.00%	149.4	194601	99.92%	99.90
Rspamd	3894	15	0.38%	19151.8	175598.6	90.17%	88.08
Spamhaus DQS	3909	0	0.00%	558	194192.4	99.71%	99.64
ZEROSPAM	3909	0	0.00%	2429.8	192303.6	98.75%	98.73
Abusix Mail Intelligence*	3908	1	0.03%	1958.8	192791.6	98.99%	98.84	N/A

*This product is a partial solution and its performance should not be compared with that of other products.
(Please refer to the text for full product names and details.)

*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 blocklists included any information on the attachments; hence its performance on the malware corpus is added purely for information.
† The standard deviation of a product is calculated using the set of its hourly spam catch rates.
(Please refer to the text for full product names and details.)

	Speed
	10%	50%	95%	98%
Axway
Bitdefender
Cleanmail Domain Gateway
Cyren eXpurgate
FortiMail
Libraesva
Rspamd
Spamhaus DQS
ZEROSPAM

(Please refer to the text for full product names and details.)

0-30 seconds

30 seconds to two minutes

two minutes to 10 minutes

more than 10 minutes

(Please refer to the text for full product names.)

(Please refer to the text for full product names and details.)

(Please refer to the text for full product names and details.)

 (Please refer to the text for full product names and details.)

Appendix: set-up, methodology and email corpora

The full VBSpam test methodology can be found at https://www.virusbulletin.com/testing/vbspam/vbspam-methodology/vbspam-methodology-ver20.

The test ran for 16 days, from 12am on 13 February to 12am on 1 March 2021 (GMT).

The test corpus consisted of 199,023 emails. 194,992 of these were spam, 31,303 of which were provided by Project Honey Pot, with the remaining 163,689 spam emails provided by Abusix. There were 3,909 legitimate emails (‘ham’) and 122 newsletters, a category that includes various kinds of commercial and non-commercial opt-in mailings.

302 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, 1,059 emails from the spam corpus were found to contain a malicious attachment while 1,835 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.

Footnotes

¹ Spamhaus DQS is a custom solution built on top of the SpamAssassin open-source anti-spam platform.

² This product is a partial solution. Such a solution has access only to part of the emails and does not receive the emails through SMTP.

³ http://www.postfix.org/XCLIENT_README.html

Latest reviews: