Install python3 and the requirements:
sudo apt install python3 python3-pip
And the required python3 packages with:
sudo pip3 install -r requirements.txt
We require energy and timing units of the system to calculate the exact leakage rates and scaling coefficients in Watt for the experiments. Therefore, please execute get_units.sh on the system where the poc is executed and enter the returned values in unit_scaling.py.
Example of the initial correlation analysis from Table 2.
python3 cli.py \
-i log.csv \
--power.init confg 1 \
-g Exp,v0,g0 \
--per REnergyPP0,RTicks 5 95 \
-u -g Exp \
--power.set_comp all \
--power.find_coefs RPowerPP0 0- The command loads the CSV file.
- The additional power-related columns are computed with scaling set to
config(i.e, use the given config file). - We group the data based on the observable variables, i.e., the Experiments, the used guesses, and the unknown but constant victim value (see threat model).
- We remove outliers that occur to scheduling etc.
- We regroup only to consider the experiments.
- We select all model components for the linear regression.
- We compute the linear regression for the PP0 power measurements.
Example output:
> --power.find_coefs column='RPowerPP0', independent_coefficients=False
rho rho_l rho_u pv hd_v0_g0 hd_v1_g1 hd_v1_v0 hd_g1_g0 hd_v1_g0 hw_g0 hw_g1 hw_v0 hw_v1 N SNR
Exp
01_n00_n64_n01_PF_NOE_DPF_DSG_D_00 0.001825 -0.00 0.00 0.16 0.000000 0.000000 0.000195 0.000066 0.000098 0.000000 0.000063 0.000337 0.000118 601135 0.000000
02_n00_n64_n01_PF_L1E_DPF_DSG_D_00 0.088806 0.09 0.09 0.00 0.003258 0.003207 0.000000 0.000000 0.000000 0.010426 0.010537 0.001472 0.001315 575321 0.678784
03_n00_n64_n01_PF_L12EOpt2_DPF_DSG_D_00 0.144881 0.14 0.15 0.00 0.001531 0.001692 0.001866 0.021410 0.000654 0.006291 0.006394 0.000000 0.000000 578844 0.203599
04_n00_n64_n01_ST_NOE_DPF_DSG_D_00 0.003509 0.00 0.01 0.01 0.000041 0.000415 0.000211 0.000264 0.000077 0.000000 0.000000 0.000393 0.000084 582063 0.004702
05_n00_n64_n01_ST_L1E_DPF_DSG_D_00 0.062798 0.06 0.07 0.00 0.001529 0.001598 0.000000 0.003959 0.000000 0.004034 0.006128 0.000474 0.000430 572005 0.259520
06_n00_n64_n01_ST_L12EOpt2_DPF_DSG_D_00 0.251175 0.25 0.25 0.00 0.003916 0.005050 0.005152 0.061032 0.002637 0.010325 0.033747 0.000772 0.005074 598650 0.506089
07_n00_n64_n01_LD_NOE_DPF_DSG_D_00 0.074373 0.07 0.08 0.00 0.013612 0.000014 0.000000 0.000187 0.000028 0.008644 0.000081 0.000459 0.000345 569656 3.962873
08_n00_n64_n01_LD_L1E_DPF_DSG_D_00 0.610248 0.61 0.61 0.00 0.017648 0.008242 0.000000 0.000000 0.000000 0.177964 0.079366 0.010929 0.006136 581152 3.726701
09_n00_n64_n01_LD_L12EOpt2_DPF_DSG_D_00 0.499520 0.50 0.50 0.00 0.008374 0.003646 0.004012 0.042198 0.003011 0.096289 0.040578 0.007548 0.004854 587375 1.654727
10_n00_n64_n01_STNT_NOE_DPF_DSG_D_00 0.001390 -0.00 0.00 0.29 0.000000 0.000167 0.000000 0.000000 0.000165 0.000098 0.000094 0.000000 0.000000 583459 0.000729
11_n00_n64_n01_STNT_L1E_DPF_DSG_D_00 0.060876 0.06 0.06 0.00 0.001346 0.001516 0.000084 0.003504 0.000284 0.003924 0.005650 0.000787 0.000835 592903 0.234321
12_n00_n64_n01_STNT_L12EOpt2_DPF_DSG_D_00 0.243347 0.24 0.25 0.00 0.003951 0.005141 0.005546 0.059606 0.002653 0.010118 0.033187 0.001741 0.003478 608419 0.513221An example of the CPA similar to Figure 10.
python3 cli.py \
-i log.csv \
--power.init config 0 \
--sel x.Id==7 \
-g v0,g0 \
--per REnergy,IEnergy 5 95 -u \
--power.cpa DPower 100 3,5,10,20,30,50,100,200,300,500,750,1000,2000,3000,5000,10000 \
--print --plot.line '1st' 1- Load the CSV.
- Init power columns.
- Select a specific experiment (with Id=7).
- Group by observable experiment.
- Remove outliers and ungroup.
- Perform the CPA on the DPower column with 100 tests each using 3,5,... etc. samples.
- Visualize the results for the percentage of the CPA with the correct value as the best candidate.
Usage: cli.py [OPTIONS] COMMAND1 [ARGS]... [COMMAND2 [ARGS]...]...
Options:
-h, --help Show this message and exit.
Commands:
--csv Usage: [OPTIONS]
> print csv of current data frame
--cut Usage: [OPTIONS] OUTLIERS LOWER UPPER
> cutaway outliers
--filter Usage: [OPTIONS] COLUMNS [mean|median]
[NUMBER_SAMPLES] > subtract moving
average filter
--head Usage: [OPTIONS]
> head of current data frame
--idx Usage: [OPTIONS] COLUMNS
> set data frame index
--merge Usage: [OPTIONS] ID_COLUMN IDS
> load intermediate data
--mvper Usage: [OPTIONS] OUTLIERS LOWER UPPER
[NUMBER_SAMPLES] > cutaway outliers
--mvstd Usage: [OPTIONS] OUTLIERS LOWER UPPER
[NUMBER_SAMPLES] > cutaway outliers
--per Usage: [OPTIONS] OUTLIERS LOWER UPPER
> remove percentile outliers
--plot.filtered Usage: [OPTIONS] COLUMN INTERPOLATE
--plot.heatmap Usage: [OPTIONS] COLUMN X Y
--plot.hist Usage: [OPTIONS] COLUMN
--plot.kde Usage: [OPTIONS] COLUMN
--plot.line Usage: [OPTIONS] COLUMN XLOG
--plot.line_xf Usage: [OPTIONS] COLUMN XLOG
--plot.overview Usage: [OPTIONS] COLUMN
--plot.scatter Usage: [OPTIONS] COLUMN
--plot.show Usage: [OPTIONS]
--plot.stats Usage: [OPTIONS] COLUMN
--power.cpa Usage: [OPTIONS] COLUMN REPETITIONS N_SAMPLES
> perform a CPA on the data column
--power.cpa_edge_cases Usage: [OPTIONS] DO_EDGE_CASES
> set of the CPA should also perform edge cases
--power.find_coefs Usage: [OPTIONS] COLUMN INDEPENDENT_COEFFICIENTS
> find the model coefficients for the model
components
--power.init Usage: [OPTIONS] [SCALE] EXPLODE
--power.set_coefs Usage: [OPTIONS] COEFFICIENTS
> set the model coefficients
--power.set_comp Usage: [OPTIONS] COMPONENTS
> set the model components
--power.set_functions Usage: [OPTIONS] [pearson|spearman] [classic|huber]
> set the statistic functions to use
--print Usage: [OPTIONS]
> print current data frame
--run Usage: [OPTIONS] CODE
> run code snippet
--sel Usage: [OPTIONS] CODE
> select subset of data
--std Usage: [OPTIONS] OUTLIERS LOWER UPPER
> remove standard outliers
-g Usage: [OPTIONS] COLUMNS
> group data
-i Usage: [OPTIONS] FILE_NAME
> load csv file
-l Usage: [OPTIONS] ID
> load intermediate data
-o Usage: [OPTIONS] FILE_NAME
> store csv file
-s Usage: [OPTIONS] ID
> save intermediate data
-u Usage: [OPTIONS]
> ungroup data