Link Search Menu Expand Document
Fault Analyzer doc.

In depth analysis

Modularity of the analyzer

When an experiment is analyzed using the analyze command, the analyzer is built to match the manip to analyze. This is done thanks to the modularity of the analyzer.

Modules

In this section we will go through the various available modules of the analyzer and when they are loaded.

Base

This module is the base of the fault analyzer, it is loaded all the time. It able to have the following information about the manip:

  • The fault general information: this result contains the number of faults, the fault probability and the number of faulted observed per fault.
  • The observed statistics: this result contains the fault probability for every observed with their name, initial values and if they are tested or not.
  • The faulted values statistics: this result contains the distribution of the faulted values and their probability of occurrence.

Fault Models

This module able the analyzer to identify various fault models. It is loaded when the type of every initial values (defined in the parameters file) is an int.

It allows to have the following information about the manip:

  • Fault model statistics: this result contains the distribution of the identified fault models and their probability of occurrence.
  • Unknown fault model values: this result contains all the faulted values for which the analyzer could not determine a fault model explaining its appearance.
  • {FAULT MODEL} destination occurrence: this result contains the probability for each register to be faulted for every {FAULT MODEL}. This is add if the corresponding {FAULT MODEL} appears at least once.
  • {FAULT MODEL} origin occurrence: in the case the determined {FAULT MODEL} involves a known value, and the origin of this value has been determined, this result contains the probability for each source to be involved in the corresponding {FAULT MODEL}. For instance, the probability for each register to be the origin of the XORed value in the case the analyzer determine the fault model is XOR with other observed.

Adding a new fault model

See how to add a new fault model.

Fault Models After Execution

This module is very similar to the Fault Models module. The only difference is that it identifies the fault models based on the observed values in the log instead of the default values given in the parameters file.

It is loaded when at list one value in the to_test parameter in the parameters file is set to False. This will force the analyzer to consider this value as a potential value involved in the fault models.

It able to have the following information about the manip:

  • Fault model statistics: this result contains the distribution of the identified fault models and their probability of occurrence.
  • Unknown fault model values: this result contains all the faulted values for which the analyzer could not determine a fault model explaining its appearance.
  • {FAULT MODEL} after execution destination occurrence: this result contains the probability for each register to be faulted for every {FAULT MODEL}. This is add if the corresponding {FAULT MODEL} appears at least once.
  • {FAULT MODEL} after execution origin occurrence: in the case the determined {FAULT MODEL} involves a known value, and the origin of this value has been determined, this result contains the probability for each source to be involved in the corresponding {FAULT MODEL}. For instance, the probability for each register to be the origin of the XORed value in the case the analyzer determine the fault model is XOR with other observed.

The Fault model statistics and Unknown fault model values results replace the ones set by the Fault Models module. However, the best thing to do would be to merge them.

Carto

This module able to analyze cartography manips and generate matrices from them. This module is loaded if the sequence "carto" is in the name of the manip file. For instance, bcm2837_aes_iv1_carto.csv will automatically load the carto module.

It able to have the following information about the manip:

Specific parameters

To work properly, the carto module requires a new parameter to be set in the parameters file:

  • coordinates_name: a list containing the names of the coordinates used in the manip file.

In some cases, the manip .csv file only stores the experiments where a fault occurs. In this case the carto module cannot parse the matrix dimension from the manip file. Therefore, to have to correct matrix size, it is possible to add the dimensions used during the experiment with the carto_resolution parameter.

  • carto_resolution: the resolution used during the experiment for the cartography. It is a couple which will be used for defining the size of all the matrices set by the carto module.

Manip and parameters file examples 

# test_base_0_carto.csv

,log,reboot,x,y
0,FlagBegin;-131071;FlagEnd,False,1,3
1,FlagBegin;-131070;FlagEnd,False,2,3

# test_base_0.py

params = {
    "log_flag_begin": "FlagBegin",
    "log_flag_end": "FlagEnd",
    "obs_names": ["My register"],
    "default_values": [0xfffe0001],
    "to_test": [True],
    "reboot_name": "reboot",
    "log_name": "log",
    "log_separator": ";",
    "nb_bits": 32,
    "coordinates_name": ["x", "y"]
}

The values corresponding to the fields mentioned by the coordinates_name parameters must contain integers. These values will be used to build the matrices.

It able to have the following information about the manip:

  • Fault matrix: the matrix giving the number of fault for every position.

Fault Model Carto

This module is a mix between the carto and the fault model modules. It generate a matrix of the positions a fault model has been observed for every fault model. To work properly, the parameter and manip files must match the requirements of the carto module.

Currently, it is automatically loaded with the carto module. However, due to its need in memory space, it might be smart to disable it by default.

The auto-saving of results was disabled because this module generated to big files.

It able to have the following information about the manip:

  • Unknown fault model matrix: the matrix giving the number of fault for which no fault model has been determined for every position.
  • {FAULT MODEL} matrix: the matrix giving the number of fault matching the {FAULT MODEL} fault model for every position.

Delay

This module able to have information about the manip when a delay parameter is set during the manip. It able to have the following information:

  • Number of faults per delay: this results stores the number of fault for every tested delays.

Specific parameters

The delay module requires specific parameters to work properly:

  • delay_name: the name of the field containing the delay information for each experiment.

This way the module will automatically extract the delay from the manip file and create the list of tested delays. However, in some cases, the manip file does not store all the tested experiments but only the ones that ended up in a fault. Therefore, it is possible to gives the module more parameters for it to be able to know all the tested delays:

  • delay_start: the first tested delay.
  • delay_end: the last tested delay.
  • delay_step: the step used to increment the tested delays.

These parameters are optional, but if they are given, the analyzer will calculate all the tested delays as the linear space defined by these parameters.

Adding a new module

See how to add a new analyzer module.