NEW AV-comparatives Real World testing results are out - (August 2013)

Lots of surprises here, Qihoo tanked, and Panda Cloud is towards the top of the pack. Kind of surprised, really. Vipre got tested again, too!

(PDF link) http://www.av-comparatives.org/whole-product-dynamic-real-world-protection-test-august-2013/

(Just the chart) http://chart.av-comparatives.org/chart1.php?chart=chart2&year=2013&month=8&sort=1&zoom=3

 

Cool.
Glad to see the 3 AV suites in my signature still holding strong in top 3.

 

Bullguard
Jerry

 

Wow, suprising.

- Bullguard, GDATA, Qihoo, and Emsisoft that low?
- AVG below MSE if user-dependent were all allow?
- Tencent near the top?

Weird month.

 

Exactly Brandonn...this month. Wait till next month. Different set of samples, different scores.

 

Also, this year seems to belong to Bitdefender, Kaspsersky, Trend Micro, and F-Secure. They are always neck and neck near the top of the chart.

 

someting strange happen, bullguard, use bitdefender engine with 92.1%??maybe a old version?
escan use bitdefender too and scores 95.1, gdata 96,f-secure 99.5...and the "real" bitdefender 100%...strange result's...

 

Can someone explain what is user dependent ? I am using AVG and bit concerned about its performance. Is it always allow or block ?

 

It asks you to block or accept. Not sure what kind of prompts that AVG have.

 

Bitdefender taking a page out of Google's play book.

 

That's why I think it is silly to look at results of just one month. I would rather see a rolling 6months (or 12).

 

I'm guessing it would have do with whether they use BD's engine, signatures, or both, and who knows how the rest of their program affects detection.

 

http://pctechmag.com/2012/08/best-anti-virus-programs/

 

Thanks aztony.

 

Anytime, friend.

 

Though being a ever-changing and questionable type stuff.

Avast did suprisingly very well,only let in 0.9% with user dependency.fortinet let in 1.6% without any user dependency factor.

Qihoo 5.2% bypassed,eset: 2.3% bypassed and Gdata WOW factor...kidding me!? I dont even believe that part of the results.Give it some air,I am not going to go with that result.Take with some salt

 

GData and Emsisoft only have 0.2% compromised, user-dependant is practically treated as the same though. Don't know what happened to Bullguard and Qihoo. Was Tencent tested with different results?

 

Although e.g. Bullguard and Qihoo use the scan engine / signatures of Bitdefender, they do not use the URL-Filter and behavior-blocker etc. of Bitdefender, due to which they can score differently.

 

I also count user dependent as detections. I never look just the green bar. So, bottom line, GDATA and even AVG did very well despite being positioned to the right side of the chart...

 

And also surprising:

- Panda, TrendMicro and Sophos so high ??

 

What happens with the user dependent stuff for AVs using another engine, like UnThreat using Vipre (which I assume is Emisoft)?

 

Always prompt, but you can tweak the settings of Identity protection (formerly Norton Antibot).

What happened to Bullgard?

 

What proves that file detection tests mean almost nothing. Wake up please.

 

Hi, you need to check the setting for identity protection component in AVG. If it is set to prompt you can change it to block automatically. You may also verify whether PUP detection is turned on or not.

 

Expected

Head of the paid pack
- Bitdefender
- Kapersky
- Trend
Head of the free friends
- Fortinet
- Avast

Movers & Shakers

Movers (down)
- Bullguard
- Eset
- Q360
- Emsisoft

Shakers (up)
- Panda
- Ahnlab (no for once better as MSE/WinDef)

WIth small sample set, average performance with variation tells something about strength (e.g. Bitdefender allways in top, Avast allways well above median)

See example of Fortinet, with calculator and spreadsheet you can determine the performance of your favourite AV (Fortinet just taken as example, since it has no user dependance). This is an accepted model for risk with small samples and few observations. The lower the deviation, the higher the problability a new observation (or test score) is relevant for the general performance. In this case in real world tests.