[BreachExchange] New Spectre Variant Hits the Network

[Destry Winant]

https://www.darkreading.com/attacks-breaches/new-spectre-variant-hits-the-network/d/d-id/1332431

A new proof of concept is a reminder that complex systems can be
vulnerable at the most basic level.

Spectre is back, and this time in a variant that adds something truly
new: remote access to cached data. The good news is that access comes
at a snail's pace.

In a research paper published last week, four researchers from Graz
University of Technology detailed NetSpectre, "a generic remote
Spectre variant 1 attack."

"In theory, it's a big deal, and they made it sound like a big deal"
says Chris Morales, head of security analytics at Vectra. "It's a
proof of concept showing it's feasible to leak information over the
network, but it's so slow it's unusable."

Just how slow is NetSpectre? In the original proof of concept, the
researchers at Graz were able to exfiltrate 15 bits per hour using a
"bit-leak gadget" that they developed for use over a network. They
were able to push data transfer up to 60 bits per hour when they
employed a "novel high-performance AVX-based covert channel."

Obviously, no actor is going to attempt to steal a database of
millions of credentials through an exploit with this performance.
"These Spectre attacks are designed to extract data from memory, which
would include user passwords and small bits of personal information,"
Morales said. "This is a reconnaissance technique, not a data
exfiltration technique."

Major Actors
In a real-world scenario, the data transfer would likely be even
slower than the proof of concept indicates. "If you're a remote
attacker, you're talking three to six bits per hour, but it's a really
low number of bits you can extract in a day," says Mounir Hahad, head
of threat research at Juniper Networks. "And since you're blind to the
memory layout of the target machine, it's going to take a long time."

The nature of NetSpectre limits its real interest to a small handful
of players, he adds. "This is beyond the capability of cybercrime.
This is for well-funded state actors who can afford to have teams go
out and find the economical applications of the technique," Hahad
explains.

In fact, Hahad predicts, "Five to 10 years from now, we'll have a leak
that says people have been using [Spectre vulnerabilities] for several
years."

Even so, both Morales and Hahad say that their researchers look at
NetSpectre and see an interesting proof of concept that has little
practical use. The reason is that other, simpler exploits are far more
productive and economical. Spear-phishing, for example, remains the
researchers' weapon of choice when it comes to extracting user
credentials from an organization. But NetSpectre is "low-level, close
to the hardware, and very complex," Hahad says. Spear-phishing is none
of those.

Warnings And Precautions
According to Morales, NetSpectre includes one obvious spot for more
development: "The bit-leak gadget is the key here," he says. "I'm sure
there's going to be more work. The first step was proving that it was
feasible."

Hahad agrees, saying that more useful gadgets may be closer than we
think. "The gadgets are out there, and the bad guys are going to find
them," he says.

In the meantime, organizations should perform the basic security steps
necessary to protect their systems from all the Spectre variants.
"Make sure the patches for Spectre are in place," Morales says. "Next,
evaluate whether you have any systems where the patch can't be
deployed."

Other security professionals echo Morales' advice. "By now,
organizations should have already taken the necessary steps to reduce
the risk of this vulnerability by patching susceptible systems,
limiting network access, and protecting privileged access to critical
systems that are still exploitable," says Joseph Carson, chief
security scientist at Thycotic.

As with many other exploits, a primary worry is systems that can't be
patched or updated because they're part of embedded process control or
dedicated application environments.

Intel- or ARM-based control systems that could remain unpatched do
have one saving grace, according to experts: They're not likely to be
host to mountains of sensitive information. Still, NetSpectre, like
the Spectres that have come before it, is a reminder that complex
systems can be vulnerable at the most basic level, and that legions of
researchers are out there eager to demonstrate just where those
vulnerabilities lie.