All Your Intel L1 Cache Belongs to CacheOut

Once again, a novel "speculative execution side-channel" attack has been discovered by researchers. They show in their paper on the attackthat it will work against modern Intel microprocessors.

Intel refers to the vulnerability as L1D Eviction Sampling, since it can enable an attacker to read from the CPU's Level 1 Data Cache. CacheOut/L1D can leak information from other processes running on the same thread, or across threads on the same CPU core.

The attack has been given the moniker of "CacheOut" by the researchers and there is a snappy website with logo that is devoted to it. As of yet, there have been no signs of this style of attack in the wild. CVE-2020-0549 has been assigned to it, with a CVSS score of 6.5 (Medium). The need for an attacker to have local access to the target for the exploit to succeed no doubt contributed to the lowering of the severity score.

The researchers say that the AMD line of microprocessors is not affected by CacheOut, since AMD does not offer any feature similar to Intel's Transactional Synchronization Extensions (TSX). TSX is there so the processor can guarantee that the instructions that form a transaction either all execute to completion or none of them will execute at all.

Arm and IBM do have a feature similar to Intel TSX, but the researchers are currently unaware of whether any of their products are affected. They are also unaware of any other attack vectors to exploit CacheOut.

The extent of the vulnerability extends across the entire chip architecture. CacheOut exploits the hardware vulnerabilities that were uncovered to dump the contents of Intel Security Guard Extensions (SGX) enclaves, for example. As such, any information stored inside the enclave can be potentially leaked by CacheOut. That's not supposed to be happening.

There are no data trails left after exploitation, as well. CacheOut does not leave any traces in traditional log files.

Virtual machines are also vulnerable to this. CacheOut exploits hardware security vulnerabilities inside the processor to leak information from both the virtual machine manager (hypervisor) and co-resident virtual machines, which is one of the things that the use of SGX was supposed to prevent.

The techniques recommended by Intel to mitigate previous similar attacks (ZombleLoad, for example) don't work in this case. This mitigation is incomplete, as the researchers showed that they can force the victim's data out of the L1-D Cache into the microarchitectural buffers after the operating system clears them. They showed that they could subsequently leak the contents of the buffers and obtain the victim's data.

Intel says it is working on microcode patches to fix the vulnerability. But the open question is how much these sort of patches -- when they come -- will affect the overall performance of the microprocessor.

— Larry Loeb has written for many of the last century's major "dead tree" computer magazines, having been, among other things, a consulting editor for BYTE magazine and senior editor for the launch of WebWeek.