Power Analysis

In our last few posts we looked at Timing Analysis and techniques such as Flush and Reload and Prime and Probe that exploit timing difference to deduce operations of victim.

Looking at Timing Analysis, we observe how we exploit the lack of uniformity within the system to observe the victim. In timing analysis, we exploit the timing difference between loading data from cache and loading data from main memory as a device to observe bits of program victim is executing. If this execution pattern depends on some secret held by the victim, we can infer details about the secret from the patter on execution of victim thread.


Alt: An icon representing power
Source: icons made by Freepik (link opens a new tab) from Flaticons (link opens a new tab)


Energy and Execution

Modern microprocessors are complex feat with multiple physical cores and multiple different functional unit - Arithmetic and Logic Unit (ALU) to process integers, Floating Point Unit, Single Instruction Multiple Data (SIMD) units such as the ones that operate via AVX Instructions in x86 architecture.

With different circuits having different complexity, they have different power draws. A careful observer with a device that can accurately measure power consumption can deduce information about runtime of victim.

Block diagram on Intel Sandy Bridge CPU showing different load store units, ALU units for integers and floating point units.

Alt: Block diagram on Intel Sandy Bridge CPU showing different load store units, ALU units for integers and floating point units.
Source (link opens a new tab): https://www.researchgate.net/figure/Sandy-Bridge-block-diagram_fig1_220308557

 

Device for leak

The device for significant leaks using power analysis are generally the ones that are dependent on SIMD circuits. Consider an algorithm that calls a SIMD routine based on value of data. For example:

while (secret)

{

        if (secret & 1)

        {

                simd_compute();

        }

        secret = secret >> 1;

}

 

Here call to simd_compute() is queued to the SIMD unit that can do 4x the computation of a Floating Point unit in same number of cycles. What this also means is that it will consume proportional energy to do so. Thus observing the power draw can give an attacker an insight of the victims runtime and sometimes  even leaking secret data.

 

How significant is power draw difference?

Using the value measured by Dr. Ian Cutress of AnandTech as reference we can see that a CPU benchmark program that stresses the Single Instruction Single Data (SISD) unit consume not more than 200W

Power draw of latest Intel 11th Generation Rocket Lake series running Agisoft, CPU power. The average power draw is close to 150W with peak nearing 200W.
Alt: Power draw of latest Intel 11th Generation Rocket Lake series running Agisoft, CPU power. The average power draw is close to 150W with peak nearing 200W.
Source (link opens a new tab): https://www.anandtech.com/show/16495/intel-rocket-lake-14nm-review-11900k-11700k-11600k/5

Compare these numbers to power draw of AVX-512 unit, we see the average close to 250W and a peak close to 300W when the SIMD unit is being used.

Alt: Power draw of latest Intel 11th Generation Rocket Lake series running 3DPMavx in AVX-512 mode. The average power draw is close to 250W with peak nearing 400W.
Source (link opens a new tab): https://www.anandtech.com/show/16495/intel-rocket-lake-14nm-review-11900k-11700k-11600k/5

As we see there is indeed a significant difference in the power draw between a normal workload and workload that depends on SIMD processing.


Complications

The following are the complications to power analysis:

  • Attacker must be able to accurately observe and record power draw of the CPU.
  • In some systems, power draw of entire system might not capture CPU state due to presence of other computation hardware.
  • Careful analysis needs to be performed on power draw data infer correct results.


We can see, as a result of the non-uniformity in power consumption by different units of CPU, the power consumption becomes a side-channel to extract information from the unsuspecting system.

 

Coming up next

 We'll take a look at a real world exploit that uses the timing based side channel and techniques around it to extract data from the operating system without need for any privilege escalation.

 

Thank you for reading till the end. I'm an undergraduate student keenly interested in Computer Architecture and I look at micro-architectural based attacks to understand more about the working of our hardware. If you find any inaccuracies in the above post, please leave a comment and I'll address it in the next edit. Have a nice day!

Comments

Post a Comment

Popular posts from this blog

Prime and Probe

Image
This is the third post in our series of Timing Analysis and I high recommend reading the first two posts - Timing Analysis (link opens a new tab), and Flush and Reload (link opens a new tab). That said if you are aware about cache hierarchy, shared libraries, shared pages, and cache placement policies, you can continue reading this article. In this post we look at Prime and Probe . Alt: An illustration of prime and probe Made with https://excalidraw.com/ (link opens a new tab) Cache Replacement Strategies The placement of data in cache is discussed briefly in Flush and Reload post. Here we will take a look at how data is replaced in cache. There are many different strategies to evict data from cache when a new data is needed to be stored in a filled cache set. Some of them are (some more practical than others): First In First Out (FIFO) - The logic behind the First In First Out is that the data that came first is probably the one that won't be used again for long time and henc
Read more

Row Hammer: Flipping Bits in Memory Without Accessing Them

Image
 This post we'll take a look at Row Hammer, a read disturbance phenomenon observed in commodity DRAM, first unearthed in the paper: Y. Kim et al., "Flipping bits in memory without accessing them: An experimental study of DRAM disturbance errors," 2014 ACM/IEEE 41st International Symposium on Computer Architecture (ISCA '14) (Link opens a new tab with PDF ~ 828KB) IEEE Page (link opens a new tab): https://ieeexplore.ieee.org/document/6853210 In this post I'll only betaking a brief look at the issue and I highly encourage you all to read the paper to find more insights and nitty gritty details of this vulnerability.   Alt: An illustration of Row Hammer showing how an activate of one row can influence the data stored in the adjacent row. The picture is made using https://excalidraw.com/ (link will open a new tab)   Structure of DRAM To fully understand the vulnerability, we must first take a look at the how modern DRAM are structured. All digital data can be represe
Read more

Speculative Execution

Image
 Most of the recent CPU vulnerability discovered depend on the property of modern processors known as Speculative Execution. In this post, we'll take a deeper look at where and why modern processors speculate. Alt: A illustration that shows speculative execution and how the results are committed or discarded based on result of speculation - whether it was correct or not respectively. The picture is made using https://excalidraw.com/ (link will open a new tab) Pipelined Execution Most modern processors are pipelined and are often superscalar in nature running instructions out of order, with multiple stages of execution of instruction running concurrently. Alt: Pipeline of a supersalar processor with 2 instruction fetch every clock cycle and a 5 stage pipeline. Source (link opens a new tab): https://en.wikipedia.org/wiki/Superscalar_processor In case of a long latency cache miss, permission check for an access or conditional branches, the processor generally has to stall waiting for
Read more