New Spectre Flaws in Intel and AMD CPUs Affect Billion of Computers

When Spectre, a class of critical vulnerabilities impacting
modern processors, was publicly revealed^[1]
in January 2018, the researchers behind the discovery said^[2], “As it is not easy to
fix, it will haunt us for quite some time,” explaining the
inspiration behind naming the speculative execution attacks.

Indeed, it’s been more than three years, and there is no end to
Spectre in sight.

A team of academics from the University of Virginia and
University of California, San Diego, have discovered a new line of attack^[3]
that bypasses all current Spectre protections built into the chips,
potentially putting almost every system — desktops, laptops, cloud
servers, and smartphones — once again at risk just as they were
three years ago.

The disclosure of Spectre and Meltdown^[4]
opened a floodgates^[5]
of sorts, what with endless^[6]
variants^[7]
of the attacks^[8]
coming to light in the intervening years, even as chipmakers like
Intel, ARM, and AMD have continually scrambled to incorporate
defenses to alleviate the vulnerabilities that permit malicious
code to read passwords, encryption keys, and other valuable
information directly from a computer’s kernel memory.

A timing side-channel attack at its core, Spectre breaks the
isolation between different applications and takes advantage of an
optimization method called speculative execution in CPU hardware
implementations to trick programs into accessing arbitrary
locations in memory and thus leak their secrets.

“A Spectre attack tricks the processor into executing
instructions along the wrong path,” the researchers said. “Even
though the processor recovers and correctly completes its task,
hackers can access confidential data while the processor is heading
the wrong way.”

The new attack method exploits what’s called a micro-operations
(aka micro-ops^[9]
or μops) cache, an on-chip component that decomposes machine
instructions into simpler commands and speeds up computing, as a
side-channel to divulge secret information. Micro-op caches have
been built into Intel-based machines manufactured since 2011.

“Intel’s suggested defense against Spectre, which is called
LFENCE, places sensitive code in a waiting area until the security
checks are executed, and only then is the sensitive code allowed to
execute,” Ashish Venkat, an assistant professor at the University
of Virginia and a co-author of the study, said. “But it turns out
the walls of this waiting area have ears, which our attack
exploits. We show how an attacker can smuggle secrets through the
micro-op cache by using it as a covert channel.”

On AMD Zen microarchitectures, the micro-ops disclosure
primitive can be exploited to achieve a covert data transmission
channel with a bandwidth of 250 Kbps with an error rate of 5.59% or
168.58 Kbps with error correction, the researchers detailed.

Intel, in its guidelines^[10] for countering timing attacks^[11] against cryptographic
implementations, recommends adhering to constant-time programming
principles, a practice that’s easier said than done, necessitating
that software changes alone cannot adequately mitigate threats
arising out of speculative execution.

The silver lining here is that exploiting Spectre
vulnerabilities is difficult. To safeguard from the new attack, the
researchers propose flushing the micro-ops cache, a technique that
offsets the performance benefits gained by using the cache in the
first place, leverage performance counters to detect anomalies in
the micro-op cache and partition the op-cache based on the level of
privilege assigned to the code and prevent unauthorized code from
gaining higher privileges.

“The micro-op cache as a side channel has several dangerous
implications,” the researchers said. “First, it bypasses all
techniques that mitigate caches as side channels. Second, these
attacks are not detected by any existing attack or malware profile.
Third, because the micro-op cache sits at the front of the
pipeline, well before execution, certain defenses that mitigate
Spectre and other transient execution attacks by restricting
speculative cache updates still remain vulnerable to micro-op cache
attacks.”