Russian
anti-virus vendor Kaspersky Lab is denying accusations that it
conducted operations designed to trick software from competing firms
into classifying harmless files as malicious.
In a story published Friday,
Reuters reporter Joseph Menn, citing two former Kaspersky Lab employees
as sources, said the company manipulated false positives off and on for
more than 10 years, with the peak period between 2009 and 2013.
Reuters
said that according to the ex-employees, Kaspersky engineers allegedly
took “important pieces of software” commonly found in PCs and injected
bad code into them so that files looked like they were infected, then
anonymously uploaded files to malware information exchange VirusTotal.
The
former employees, who wished to remain anonymous, told Reuters that
some of the attacks were ordered by company chief Eugene Kaspersky,
partly to retaliate against smaller rivals that he felt were aping his
software instead of developing their own technology.
"Eugene considered this stealing," one of the former employees said.
Kaspersky
Lab fired back immediately, saying the allegations made are not true,
and that such actions are unethical, dishonest and illegal.
“Kaspersky Lab has never conducted any secret campaign to trick
competitors into generating false positives to damage their market
standing,” Kaspersky Lab said in a statement provided to NgCyberTechTips.
“Accusations by anonymous, disgruntled ex-employees that Kaspersky Lab,
or its CEO, was involved in these incidents are meritless and simply
false,” the statement continued. “As a member of the security community,
we share our threat intelligence data and IOCs on advanced threat
actors with other vendors, and we also receive and analyze threat data
provided by others.”
While
the Russian security firm defended against the accusations, the company
did say that it conducted a one-time experiment in 2010 by uploading
samples of non-malicious files to VirusTotal—something that had been made public by the company soon after—and well before being accused in the report from Reuters.
“In
2010, we conducted a one-time experiment uploading only 20 samples of
non-malicious files to the VirusTotal multi-scanner, which would not
cause false positives as these files were absolutely clean, useless and
harmless,” Kaspersky Lab explained. “After the experiment, we made it public
and provided all the samples used to the media so they could test it
for themselves. We conducted the experiment to draw the security
community’s attention to the problem of insufficiency of multi-scanner
based detection when files are blocked only because other vendors
detected them as being malicious, without actual examination of the file
activity (behavior).”
The company also claimed that in 2012 it was affected by an unknown
source uploading bad files to VirusTotal, which led to a number of
incidents with false-positive detections.
During the VB Conference in Berlin in October 2013, Kaspersky said
there was a private meeting between antivirus vendors to exchange the
information about the incidents, work out the motives behind the attack
and develop an action plan.
“We had investigated these attacks but could not find out who was
behind them. We had some suspects, Kaspersky was not one of them,” Liam
O'Murchu, Manager of Security Response Operations for North America at
Symantec, wrote on Twitter.
The
tactic, not matter who is behind it, is a major wakeup call to the
industry, says Rahul Kashyap, Chief Security Architect and SVP of
Security & Solutions Engineering at Bromium.
“Besides the huge impact of the claim, there are two other issues this
report brings out – the challenges of reliably attributing and the
fragility of anti-virus ‘system’," Kashyap told SecurityWeek.
“To
prove that this story is indeed true, reliable facts need to be
presented that provide legit evidence against Kaspersky. I doubt it’ll
be easy for anyone to reliably attribute the act directly to
Kaspersky—unless the informants did it themselves and stored reliable
evidence at the time of crime,” Kashyap said.
“This
also exposes to the fragility of the entire malware sample distribution
system. As the report claims – a hole in the system was uncovered and
plugged after large scale damage was observed. The entire Anti-Virus
industry is about reacting after damage, this act further proves yet
another flaw in the model.”
“On the incident in general, we’re well aware of the issue of false
positives propagating from product to product through the vital
industry-wide information sharing systems,” John Hawes, Chief of
Operations at Virus Bulletin, told SecurityWeek. “In the past we
regularly observed false alarms which cropped up in our testing
spreading to other products very quickly. This was a symptom of
automated processes relying too heavily on detection results from other
vendors as a criteria for classifying an item as malicious, and it seems
to have reduced fairly considerably in the last year or two as
automation systems have become more sophisticated.”
“As to whether a security vendor would be involved in deliberately
trying to abuse these processes to discredit their rivals, I remain
highly skeptical,” Hawes said. “The benefits would be rather marginal –
to achieve a major headline-making false positive event of the kind
which would dent a vendor’s reputation, a seriously major file which
broke a major package or stopped Windows from functioning properly would
have to be made to look malicious.”
VirusTotal, a subsidiary of Google, is an aggregator and does not take
responsibility for such false positives, and has a policy to not
whitelist any files or URLs. Furthermore, the company says it will not
remove any detections resulting from the normal operation of the
products it makes use of.
Earlier this year VirusTotal launched an initiative
designed to address the issue of false positives through a project
called “trusted source.” As part of the initiative, major software
vendors are asked to share metadata from their software collection.
Additionally, when files get distributed to antivirus vendors under the
trusted source initiative, they are tagged so that potential erroneous
flags can be ignored, preventing a snowball effect with detection
ratios.
“Any major vendor with proper QA processes should be routinely running
their products over in-house sets of major files before releasing
updated detection routines, to watch out for just this sort of issue,”
Hawes said.
“Although
the security market is very competitive, trusted threat data exchange
is a critical part of the overall security of the entire IT ecosystem,
and we fight hard to help ensure that this exchange is not compromised
or corrupted,” Kaspersky stated.
ConversionConversion EmoticonEmoticon