bham.leakiest

Class CompositionalEstimate

java.lang.Object

bham.leakiest.CompositionalEstimate

public class CompositionalEstimate
extends java.lang.Object

This is a library of useful compositional reasoning about estimation of information leakage measures.
The theories used in the class appears in a paper iin QEST 2014

Version:

1.3

Author:

Yusuke Kawamoto

Constructor Summary

Constructors

Constructor and Description
CompositionalEstimate()

Method Summary

Modifier and Type	Method and Description
static double[]	estimateParallelGLeakWithSharedInput(ProbDist jpd, ProbDist apd, Channel[] channels, GainFunction gf, java.util.Set<java.lang.String> guessDomain) Returns an upper bound on the g-leakage of the channel composed in parallel in the case input is shared among the channels.
static double[]	estimateParallelMinCapacityWithSharedInput(Channel[] channels) Returns an upper bound on the min-capacity of the channel composed in parallel in the case input is shared among the channels.
static double[]	estimateParallelMinEntropyLeak(ProbDist jpd, ProbDist apd, Channel[] channels) Returns a lower and an upper bound on the min-entropy leakage of the channel composed in parallel where inputs to different channels are drawn independently.
static double[]	estimateParallelMinEntropyLeakWithSharedInput(ProbDist jpd, ProbDist apd, Channel[] channels) Returns an upper bound on the min-entropy leakage of the channel composed in parallel in the case input is shared among the channels.
static double	exactParallelMinEntropyLeak(ProbDist jpd, Channel[] channels)
static double	exactParallelMinEntropyLeakWithSharedInput(ProbDist jpd, Channel[] channels)
static double	HgMin(ProbDist pd, GainFunction gf, java.util.Set<java.lang.String> guessDomain) Returns Hg^min(pd) = - log min { pd[x] g(w, x) | x in X, w in W, pd[x] g(w, x) != 0 }.
static double	HMinInf(ProbDist pd) Returns H^min(pd) = - log min { pd[x] | x in support(pd) }.
static double[]	MInf(ProbDist jpd, int numChannels) Returns MInf defined in our compositionality paper.
static void	printEstimatedMeasure(int taskType, ProbDist[] pds, Channel[] channels, int numChannels, int sampleSize, boolean priorShared, GainFunction gf, java.util.Set<java.lang.String> guessDomain, boolean compositionalEstimate, double approxPriorLevel, boolean approxDoNotKnowChannels) Chooses to estimate and print one of leakage measures by compositional reasoning in the case of discrete inputs.
static void	printExactDiscreteMinEntropyLeakOnly(ProbDist[] pds, Channel[] channels, boolean priorShared)

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

CompositionalEstimate

public CompositionalEstimate()

Method Detail

estimateParallelGLeakWithSharedInput

public static double[] estimateParallelGLeakWithSharedInput(ProbDist jpd,
                                                            ProbDist apd,
                                                            Channel[] channels,
                                                            GainFunction gf,
                                                            java.util.Set<java.lang.String> guessDomain)

Returns an upper bound on the g-leakage of the channel composed in parallel in the case input is shared among the channels.

Parameters:

jpd - joint input probability distribution

apd - approximate joint input probability distribution

channels - array of channels

gf - gain function

guessDomain - the set of all guesses

Returns:

an upper bound on the g-leakage of the channel composed in parallel

estimateParallelMinCapacityWithSharedInput

public static double[] estimateParallelMinCapacityWithSharedInput(Channel[] channels)

Returns an upper bound on the min-capacity of the channel composed in parallel in the case input is shared among the channels.

Parameters:

channels - array of channels

Returns:

an upper bound on the min-capacity of the channel composed in parallel

estimateParallelMinEntropyLeak

public static double[] estimateParallelMinEntropyLeak(ProbDist jpd,
                                                      ProbDist apd,
                                                      Channel[] channels)

Returns a lower and an upper bound on the min-entropy leakage of the channel composed in parallel where inputs to different channels are drawn independently.

Parameters:

jpd - joint input probability distribution

apd - approximate joint input probability distribution

channels - array of channels

Returns:

a lower and an upper bound on the min-entropy leakage of the channel composed in parallel

estimateParallelMinEntropyLeakWithSharedInput

public static double[] estimateParallelMinEntropyLeakWithSharedInput(ProbDist jpd,
                                                                     ProbDist apd,
                                                                     Channel[] channels)

Returns an upper bound on the min-entropy leakage of the channel composed in parallel in the case input is shared among the channels.

Parameters:

jpd - joint input probability distribution

apd - approximate joint input probability distribution

channels - array of channels

Returns:

an upper bound on the min-entropy leakage of the channel composed in parallel

exactParallelMinEntropyLeak

public static double exactParallelMinEntropyLeak(ProbDist jpd,
                                                 Channel[] channels)

Parameters:

jpd - joint input probability distribution

channels - array of channels

Returns:

the exact value of the min-entropy leakage of the channel composed in parallel

exactParallelMinEntropyLeakWithSharedInput

public static double exactParallelMinEntropyLeakWithSharedInput(ProbDist jpd,
                                                                Channel[] channels)

Parameters:

jpd - joint input probability distribution

channels - array of channels

Returns:

the exact value of the min-entropy leakage of the channel composed in parallel in the case input values are shared among channels

HgMin

public static double HgMin(ProbDist pd,
                           GainFunction gf,
                           java.util.Set<java.lang.String> guessDomain)

Returns Hg^min(pd) = - log min { pd[x] g(w, x) | x in X, w in W, pd[x] g(w, x) != 0 }.

Parameters:

pd - input probability distribution

gf - gain function

guessDomain - the set of all guesses

Returns:

Hg^min(pd)

HMinInf

public static double HMinInf(ProbDist pd)

Returns H^min(pd) = - log min { pd[x] | x in support(pd) }.

Parameters:

pd - input probability distribution

Returns:

H^min(pd)

MInf

public static double[] MInf(ProbDist jpd,
                            int numChannels)

Returns MInf defined in our compositionality paper.

Parameters:

jpd - joint input probability distribution

numChannels - the number of channel compositions

Returns:

MInf

printEstimatedMeasure

public static void printEstimatedMeasure(int taskType,
                                         ProbDist[] pds,
                                         Channel[] channels,
                                         int numChannels,
                                         int sampleSize,
                                         boolean priorShared,
                                         GainFunction gf,
                                         java.util.Set<java.lang.String> guessDomain,
                                         boolean compositionalEstimate,
                                         double approxPriorLevel,
                                         boolean approxDoNotKnowChannels)

Chooses to estimate and print one of leakage measures by compositional reasoning in the case of discrete inputs.

Parameters:

taskType - type of calculation of a leakage measure

pds - array of probability distributions

channels - array of channels

numChannels - the number of channels

sampleSize - the number of samples

gf - gain function

guessDomain - the set of all guesses

priorShared - whether a (prior) input value is shared among all channels

compositionalEstimate - whether we use the compositional reasoning

approxPriorLevel - the summation of all probabilities removed from the input distribution

approxDoNotKnowChannels - whether the analyzer has no knowledge on channel matrices or not

printExactDiscreteMinEntropyLeakOnly

public static void printExactDiscreteMinEntropyLeakOnly(ProbDist[] pds,
                                                        Channel[] channels,
                                                        boolean priorShared)