Protecting over 200 million PCs, Macs, & Mobiles – more than any other antivirus

April 22nd, 2013

Is one search engine safer than another?

Question of the week: Since I have been using avast! I have been conscious of staying secure online. Does it matter which search engine I use? Is one safer than the other?

Thanks for using avast! to protect your computer. Yours is a great question, but maybe not one that people consider when thinking about the security of their system.

A recent 18-month study by the German Security firm AV-TEST Institute revealed that search results about breaking news stories, like the recent bombing at the Boston Marathon, frequently contain malicious links. People seek news quickly and they click on the links at the top of results without stopping to consider their safety or reliability. PCs without reliable security software soon become infected.

Google search is safer than Bing

AV-TEST evaluated about 40 million websites and found that Google is the safest way to search if you want to avoid malware. It’s rival Bing delivered five times more malicious websites in search results than Google. Yandex, Russia’s popular search engine, performed even worse than Bing delivering 10 times as many infected websites as compared to Google. This chart shows the results from Yandex, Bing, Google, and Blekko.

Overall the number of infected websites represent a small overall percentage of search results. But you need to remember that Google handles 2 to 3 billion search requests worldwide every day. The editors of the study point out, “If this total is factored into the calculations, the total number of websites containing malware found by the search engine is enough to make your head spin!”

Even though the study indicates that Google is the safer bet, all the search engines are pretty safe. As a conscientious user, you just need to be careful what you click on and make sure your programs and applications are up-to-date. Of course, you are already covered by terrific protection – avast! Antivirus! ;-)

Additional information:

Microsoft response to AV-Test- “We show results with warnings for about 0.04% of all searches, meaning about 1 in 2,500 search result pages will have a result with a malware warning on it.  Of those, only a small proportion of malicious links ever get clicked and the warning therefore triggered, so a user will see the warning only 1 in every 10,000 searches. In any case, the overall scale of the problem is very small.”

Yandex response to AV-Test – “Yandex uses its own proprietary antivirus technology to protect users from malicious software,” reads an email from the company. “Yandex marks the infected webpages in its search results in order to notify users of unsafe content. We just notify users of possible consequences and do not block access to the webpage completely.”

Categories: Uncategorized Tags:
  • tbonky

    Blekko is reported as the most spam free.
    My experience is that it acts as a browser hijacker and spam creator. Once installed, it is fairly difficult to remove. That is the first red flag for me. I would never recommend this to a computer user.

  • avastxj

    Great research / TERRIBLE GRAPH.

    The graph is misleading. To the casual reader -probably the majority of the public – it would appear that Blekko is the safest search engine. The graph is visually MEANINGLESS AND MISLEADING. The Blekko percentage is actually quite high – returning 203 malware sites for only 3 million websites. The increasing scale on the left side even further muddles things by obscuring the big difference in number of sites searched.

    The real comparison is the PERCENTAGE OF malware sites shown per sites searched on a search. The graph, because each search engine test had a differing number of searches does NOT make that clear.

    Sorry, but this is really shockingly amateur for something so well researched to be presented so poorly. Please replace the graph with a useful one that visually shows the results of the research.

  • Local

    Intuitively, you’d think that with Google’s resources and concern for their “user’s experience” that they would indeed be the most spam free. But I agree about the graph being misleading. Since the baseline of websites tested is not equal for each search engine…a percentage figure is the only real basis for comparison.

  • Hazel

    I don’t know about the accuracy of the chart – not having a mathematical mind. But I do know I get better search results from Bing. I suspect Google is too busy selling status. Sometimes the results they bring up have no relationship at all to what I’m looking for.

  • jrolandcole


    It’s bad enough that the distances between each line, starting with the first line (which means 100) are 90 % ratios of the amount represented by the next higher line. (That is to say, while lines 1, 2, 3, 4, and 5, respectively have values of 100, 1000, 10,000, 100,000 and 10,000,000, the first space represents 900, the second 9, 000, the third 90, 000, the fourth 900, 000, and the fifth 9, 000, 000, nine million!) How many even notice this unusual weirdness. And what a small per centage of those will actually be able to understand such a graph! If it were, in fact, accurate, it would still be VERY MISLEADING AND DECEPTIVE FOR THOSE REASONS ALONE.

    I had a brief impulse to be sarcastic and say, “Thanks a lot, Deborah Salmi, FOR SUCH DECEIT. Then I remember what I discovered when I taught philosophy and ethics.Text book authors rarely have anything to do with the graphics in their textbooks. The publishers or contract employees, who have abysmal to no understanding of the text, create the graphics which are supposed to make things clearer quickly. Often, instead, they often actually contradict the actual fine words and wonderful content of the authors. I suspect, here, the graphics may have a worse motivation or origin than ignorance.

    What does a comparison of Blekko with Google show? Who knows how to read such increasing scales? None but the most avid mathematicians or physicists! A normal user of a graph would say, Blekko has about 60 % of the red of Google’s amount of malware (272 cases) portrayed in the graph, which would be 163 instead of the actual 207 cases.

    203 divided by 272 = 74.6 % = Blekko’s actual percentage of malware compared to Google’s re the numbers given. But, since Blekko only had 3,007,975 cases tested compared to Google’s 10,912,207 (over three–3x+–times as many!), then
    10,912,207divided by 3,007,975 =3.6278, the ratio of Google to Blekko cases tested.. So 74.6% times that ratio (3.6278), gives the actual per centage of malware cases of Blekko compared to Google—which, turns out to be 270.630 or 271 %! That’s an actual figure that very few people could manage to get from such a graph. Therefore, I suggest the graph is (1) useless, or (2) misleading, or (3) worse.

    I agree with avastxj (April 23rd)’s call: “Please replace the graph with a useful one that visually shows the results of the research.” Avastxj accepts that a fine study was performed.

    I note in the article that Google handles some 2 to 3 Billions cases per day and that the study was done over 18 months. 18 months X an average of 30 days per month – 540 days X an average of 2.5 billion cases per day (2,500,000,000) = 1,350 billions of cases in the 18 months. Since only 10,912,207 (i.e., less than 11 millions of the 1,350 billions or 1,350,000 millions of cases) were actually tested, that means approximately1 out of every 123,714 cases were actually tested! One would have to know some high-powered statistics and/or be very smart about the actual sampling techniques used to know if representative results are in fact being reported by the article across the board.

    Microsoft reports .04 % or 1 of every 2500 search pages has malware on them and 1 of every 10, 000 searches sees a malware warning (since many options are not clicked). A similar proportion for Google would indicate that of approximately 2.5 billions searches per day, 1/2500th or exactly 1 million cases of malware per day would show up and perhaps only 1/4th of that number, or 250,000, would actually see a malware warning. (Please note that this is a per day amount or estimate based on Microsoft’s admission addeded to the article, and they consider ‘the overall scale of the problem” to be “very small.”)

    IF the same proportion holds for Google, 250,000 malware warnings per day times 540 days = 135,000,000 = 135 million malware warnings would be issued in 18 months, compared to the 272 cases of malware found, according to the numbers reported by the article.

    I will hold those estimates to be true unless someone shows that my numbers are wrong or Google explains how and why it gets vastly superior results compared to Microsoft, i.e. how exactly it gets 496,323 times better results than Microsoft does (i.e., 272 cases of malware found compared to 135 million malware warnings issued, proportionally).

  • jrolandcole

    Dear Deborah,

    Thanks for the note. If you read my comment, you will read that I assumed YOU did not create the graph. Your name, however, indicates that you wrote the article for avast, apparently. As such, you and they have some responsibility for dispensing truth or propaganda to us.

    I wish you or some qualified person would address the issues I raised above, including the apparently radical discrepancies I believe I “smoked out” and explicated. I, truly, would loved to be proved wrong.

    And we, your readers, are still clamoring for a strait-forward graph that yields readily seen and understood facts and comparisons. Can you get that for us from “the German Security firm Av-Test Institute,” and print it here instead of giving us a link to the same old same old? Will you?

    Finally, if there is a relationship between the Av-Test Institute and avast, and between Google and avast, please inform us all. From the tenor of the article making Google look “good,” I would assume there is some relationship between Google and avast. “Inquiring minds want to know.”

    You seem to be a caring person and a good reporter. Please help us find the answers. Can you? And, will You?

    Thanks. PS. I’m very pleased with what I know and have experienced of avast so far.

  • nickday30

    although the graph cannot be used to compare data it still represents it in an organized matter…its not that hard to find the percent

  • RLLBcheese

    With regards to your numbers, they all seem to be accurate except the proportion at the end. I’m not quite sure what you’re trying to get at, but with the numbers you have, the following can be shown:

    -av-test hit 272 out of the 135 million total google warnings (= 1 out of every 496323 warnings)
    -you state that google has 496323 times better figures than bing (in the number of avast hits compared to the total warnings in the 18 month period)
    -hence, that implies that av-test, in the 18 month period, received EVERY SINGLE MALWARE WARNING that bing sent out over that period.

    The above seems rather illogical to me, so I can only presume that you used the numbers for the wrong purposes, getting their meanings confused.

    According to SEOdesk (July 2011), bing gets 57 million searches per day, and 1 out of every 10,000 searches issues a warning, according to microsoft, meaning that, per day, bing will issue approximately 5700 malware warnings per day. Av-test hit 1285 over 540 days, meaning 2.38 per day. This gives a ratio of 1 out of every 2395 malware warnings hit by av-test. Also, with 11 million searches over 540 days = 20370 per day, and 57 million searches handled by bing each day, av-test performed 1 out of every 2798 searches on bing

    As for google, using the SEOdesk numbers (to keep things consistent), av-test performed 1 out of every 49091 searches on google, received 0.5 warnings per day and hit 1 out of every 200,000 malware warnings issued by google

    We divide the ratio of hits/total warnings by the proportion of searches made by av-test to normalise the data (make it such that, in terms of searches, google and bing are on level pegging, and hence we can compare the 2 of them), giving values of 1.168267 for bing and 0.245455 for google. (this number represents the ratio of the number of hits av-test got compared to what we would expect it to get. This means that, with google, av-test hit significantly less than we would expect for the number of pages visited and bing hit a few more. There are 3 possible reasons for this:

    1.) The numbers are inaccurate for google: We presumed that the proportion of pages infected with malware on google is the same as bing. However, if google is a better search engine, then it will filter out more malware-infected pages and hence the percentage will be lower, causing less hits than we would expect.
    2.) Google is very good at keeping malware out of it’s top results: The study, if you read the av-test document, looks at the top results for a search. If google is much better at keeping malware-infected pages out of the top results, then this would also account for the significantly lower number.
    3.) The search was biased: If the search had some sorts of bias, whether it be through different search terms being used, search terms known to be looked upon more favourably by a certain search engine, or just plain fabrication of the results, then that would also account for the difference.

    If reasons 1 or 2 of the above are true, or both, then it shows that, for searching for news stories, where you’re likely to just click on the very first link, google is a much safer option than bing. If 1 is correct, then seaching with google is safer regardless of which link you click on. For results later down the page, in the case of 2, a more complex analysis would be needed to work out which, if any, is better.

    With regards to the graph, it is a logarithmic scale, something which is used quite commonly in situations where data points very more evenly by orders of magnitude than by actual value. In this case, it has been used so that the axis for the pages tested can be the same axis as the one for the malware hits. The graph would be better made if it were to have 2 scales on the left axis, or an axis with it’s own scale on the right, and all scales being linear, however, that is not always possible with graph-drawing software.

    As for whether the sampling produced valid results with the sampling frequency, it is impossible to tell just from the ratio/numbers. If we had some sort of variance/standard deviation and some ranges, then it would be possible to tell.

  • G-Rex

    Has anybody also considered that though there are 2 to 3 billion search requests on google, they are not all unique? Another problem is that some of the search requests return the same or almost the same results because of related and filtered words. Words like “the”, “or”, and “a” as well as every other common English word that don’t contribute to content of the message are filtered out to narrow down search results. Google also offers top searches, which are usually clicked on by the people searching, that cuts down on some of those search results. Websites that get a lot of hits like Wikipedia and Youtube are moved to the top of your search results, and AVAST as well as many others say that most malware comes from legitimate sites. The first links are probably good for getting reliable results, but they are more likely to have malware because the site is a legitimate, frequently-used site.

    In the end, Google and other search engines don’t control what’s on the site that comes up when you click on the link, so there’s not that much difference between the engines except what they may try to do to prevent you from accessing sites that are known to have malware. If you are clicking on a page, you should check the domain first and understand that every website can be dangerous.

    @jrolandcole. You have pointed out a lot that IS wrong with the data, but you also forgot the small section of the article that says explicitly “If this total” total being the 2 to 3 billion searches “is factored into the calculations, the total number of websites containing malware found by the search engine is enough to make your head spin!”

  • maschang

    because the search engine is a program that is used as a tool to find any maca forms of information on the internet, it is very important to avoid malware and avast can make a plugin to be able to reduce risk.