qif

qif is a library for Quantitative Information Flow: it computes vulnerability and leakage measures of channels, constructs optimal privacy mechanisms, and checks refinement relations between channels. It is written in C++ (libqif) with Python bindings; this documentation covers the Python package.

Install

pip install qif

  • Needs a haswell or later CPU (AVX2)

  • Python 3.11 to 3.15

  • MacOS 15 or later (x86_64 and arm64)

To use libqif from C++ see the installation instructions.

A sample program

from qif import *

def compute_bayes(C):
    pi = probab.uniform(C.shape[0])

    print("Channel:\n", C)
    print("Prior:\n", pi)

    print("Bayes vulnerability", measure.bayes_vuln.posterior(pi, C))
    print("Bayes mult-capacity", measure.bayes_vuln.mult_capacity(C))

compute_bayes(channel.randu(5))

# same, but using rational arithmetic
set_default_type(rat)

C = numpy.array([
    [rat(1,2), rat(1,4), rat(1,4)],
    [rat(1,6), rat(3,6), rat(2,6)],
    [rat(1,2), rat(1,2), rat(0)],
])
compute_bayes(C)

Modules