New ‘Prime’ Meltdown, Spectre exploits outlined by Nvidia, Princeton University

Just one month after researchers exposed methods to extract sensitive data from a device’s memory through all modern processors, another research paper arrives to illustrate how the processor design flaw can be used in other attacks. The paper, dubbing the new exploits MeltdownPrime and SpectrePrime, derives from three researchers who work at at Princeton University and graphics chip manufacturer Nvidia. 

As reported last month, all processors dating back to at least 2011 have a flaw in the way they’re designed. Part of a processor’s speed comes from its ability to predict where the current list of instructions will go — they have “branch prediction units” that take an educated guess about what command will come next. To make these predictions, processors toss data back and forth from two memory sets: local on-chip memory called cache for fast access, and the PC’s system memory. This data isn’t secured, and that’s where the original Meltdown and Spectre attacks come in. 

The Meltdown approach applies to Intel and Apple processors. A hacker can create a malicious program to access that raw information, which could include usernames, passwords, credit card numbers, and so on. It taps into the privileged information typically only accessible by the root of an operating system, otherwise known as the kernel. 

Meanwhile, Spectre applies to Intel, AMD, and all mobile chips based on ARM’s processor design, including Apple. Here hackers can create a program to trick the processor into executing instructions not built into legitimate programs and apps installed on the PC. In other words, your favorite apps and programs could be tricked into coughing up your sensitive data. 

Both methods are merely proof-of-concepts reported by Google Project Zero, and researchers from Cerberus Technology and various universities. Both are called side-channel attacks as they don’t target specific software, such as Adobe Flash. 

The new MeltdownPrime and SpectrePrime exploits rely on an attack called Prime+Probe that takes advantage of processor “cache invalidations,” which is a method of replacing or removing entries in the CPU’s cache. Whereas Meltdown and Spectre simply “pollute” this cache during the CPU’s path prediction (aka speculative execution), the new exploits take a different approach. 

“MeltdownPrime and SpectrePrime are caused by write requests being sent out speculatively in a system that uses an invalidation-based coherence protocol,” the paper states. A coherence protocol means that the PC is keeping all data stored in cache and memory consistent. But that protocol may “invalidate cache lines in sharer cores as a result of a speculative write access request even if the operation is eventually squashed.” 

The researchers validated their findings using a MacBook packing an Intel Core i7 processor, and MacOS Sierra v10.12.6. They ran the exploit 100 times on the machine, with a 99.95-percent success rate for SpectrePrime versus the 97.9-percent rate seen with the vanilla Spectre exploit. 

“We believe that any software techniques that mitigate Meltdown and Spectre will also be sufficient to mitigate MeltdownPrime and SpectrePrime. On the other hand, we believe that microarchitectural mitigation of our Prime variants will require new considerations,” the paper states.