mirror of
https://github.com/RIOT-OS/RIOT.git
synced 2024-12-29 04:50:03 +01:00
62 lines
2.5 KiB
Markdown
62 lines
2.5 KiB
Markdown
|
# About
|
||
|
|
This test compiles and runs entropy sources. The first `zero entropy` source can be run on
|
||
|
|
`native` and does not provide real entropy values, though, it tests execution of the main
|
||
|
|
module and its common components including the optional health tests. Additional sources are
|
||
|
|
requested subsequently. As indicated in the documentation of the [entropy module](../../sys/entropy_source/doc.txt),
|
||
|
|
entropy is vulnerable and specific hardware sources require a priori validation.
|
||
|
|
The `ADC noise` source requires the `periph_adc` feature of a board. Its properties can vary widely as
|
||
|
|
depicted in the documentation of the [`ADC noise`](../../sys/include/entropy_source/adc_noise.h)
|
||
|
|
modules and proper testing and parametrization need to take place for every single platform and ideally
|
||
|
|
environmental properties before deployment. The test simply initializes the ADC noise entropy source
|
||
|
|
with the default ADC pin, requests and dumps many samples with enabled health tests of the source and
|
||
|
|
conditioning.
|
||
|
|
|
||
|
|
# Expected results
|
||
|
|
The `zero entropy` is expected to indicate different errors after requesting more samples than
|
||
|
|
the cutoff value of the Repetition Count Test (NIST SP 800-90B 4.4.1) and the Adaptive
|
||
|
|
Proportion Test (NIST SP 800-90B 4.4.2). The expected output is:
|
||
|
|
|
||
|
|
## native
|
||
|
|
|
||
|
|
```
|
||
|
|
# main(): This is RIOT! (Version: <version>)
|
||
|
|
# Zero entropy single request 0/311 returned: 0
|
||
|
|
...
|
||
|
|
# Zero entropy single request 20/311 returned: -3
|
||
|
|
...
|
||
|
|
# Zero entropy single request 310/311 returned: -3
|
||
|
|
# Zero entropy single request 311/311 returned: -5
|
||
|
|
# Zero entropy request 64 Bytes: -5
|
||
|
|
|
||
|
|
```
|
||
|
|
|
||
|
|
## Board
|
||
|
|
|
||
|
|
The `ADC noise` source test should dump unpredictable values. The *von Neumann* conditioning internally
|
||
|
|
requests multiple samples so that subsequent values include bit changes. Thus, it has
|
||
|
|
a variable runtime. To explore this, one test requests a buffer with entropy values
|
||
|
|
and it measures the processing time. If no changes are in place, the conditioning function
|
||
|
|
will abort after exceeding a threshold.
|
||
|
|
Otherwise, no health test errors should be indicated. Please note that
|
||
|
|
missing errors do **not** reveal information about the quality of an entropy source.
|
||
|
|
|
||
|
|
```
|
||
|
|
# main(): This is RIOT! (Version: <version>)
|
||
|
|
# Zero entropy single request 0/311 returned: 0
|
||
|
|
...
|
||
|
|
# Zero entropy single request 20/311 returned: -3
|
||
|
|
...
|
||
|
|
# Zero entropy single request 311/311 returned: -5
|
||
|
|
# Zero entropy request 64 Bytes: -5
|
||
|
|
# ADC noise source entropy/sample: 65536 [2^16 * bit / sample]
|
||
|
|
# ADC noise source entropy/sample: 1 [bit / sample]
|
||
|
|
# 95
|
||
|
|
# 9f
|
||
|
|
# a5
|
||
|
|
# 01
|
||
|
|
# e1
|
||
|
|
# 3e
|
||
|
|
# 73
|
||
|
|
...
|
||
|
|
# ADC noise request 64 Bytes returned: 0. Time: <time>us
|
||
|
|
```
|