Specter and Meltdown, names given to a recently discovered vulnerability that affects almost every computer chip manufactured in the last 20 years. If exploited, attackers could gain access to data previously considered completely protected. The Specter and Meltdown flaws work by exploiting two important techniques used to make CPU chips execute faster, called speculative execution and caching.
Speculative execution allows a CPU to attempt to predict the future to work faster. For example, if the chip determines that a program contains multiple logical branches, it will start calculating the values for all of the branches before the program decides which branch to take. When the correct branch is determined, the CPU has already produced the values for that branch. If the CPU sees that the same function is frequently used, it might use idle time to compute that function so it has what it thinks the answer will be ready if needed.
Caching is used to speed up memory access. Random access memory (RAM) is located on separate chips and it takes a relatively long time for the CPU to access data in the RAM. There is a special small amount of memory storage called CPU cache that is built on the CPU chip itself that can be accessed very quickly. This cache memory gets filled with data that the CPU will need soon or often. Data that is produced by such speculative execution is often stored in the cache, which contributes to making it a speed booster. The problem arises when caching and speculative execution start circumventing protected memory.
Protected memory is a foundational concept underlying computer security. It allows a program to keep some of its data private from some of its users, and allows the operating system to prevent one program from seeing data belonging to another. In order to access data, a process needs to undergo a privilege check, which determines whether or not it’s allowed to see that data.
A privilege check can take a relatively long time. Due to speculative execution, while the CPU is waiting to find out if a process is allowed to access that data, it starts working with that data even before it receives permission to do so. The problem arises because the protected data is stored in CPU cache even if the process never receives permission to access it. Because CPU cache memory can be accessed more quickly than regular memory and due to the long latency associated with privilege checks, the process can potentially access certain memory locations that it shouldn’t be allowed to access. As this problem exists in the hardware there is no direct way to correct it. Software patches have been offered to mitigate the exposure but have led to some degradation in performance of the CPU. In many cases, the software patch is targeted at a specific product and installing the wrong patch can severely impact system operation.
The most immediate action security teams and users can take to protect computer systems is to prevent execution of unauthorized software and avoid access to untrusted websites. Security policies must be are in place to prevent unauthorized access to systems and the introduction of unapproved software or software updates.
*Prof. Bill Pierce, the author of this article, is an Assistant Professor of computer science at the Department of Computer Science & Information Technology at Hood College in Frederick, Maryland. He teaches undergraduate and graduate courses in Computer Architecture, Digital Logic and Switching Theory, Digital Signal Processing and Musical Computing.*