tutorial.mpc

# sint: secret integers

# you can assign public numbers to sint

a = sint(1)
b = sint(2)

def test(actual, expected):

    # you can reveal a number in order to print it

    actual = actual.reveal()
    print_ln('expected %s, got %s', expected, actual)

# private inputs are read from Player-Data/Input-P<i>-0
# or from standard input if using command-line option -I

for i in 0, 1:
    print_ln('got %s from player %s', sint.get_input_from(i).reveal(), i)

# some arithmetic works as expected

test(a + b, 3)
test(a * b, 2)
test(a - b, -1)

# Division can mean different things in different domains
# and there has be a specified bit length in some,
# so we use int_div() for integer division.
# k-bit division requires 4k-bit computation.

test(b.int_div(a, 16), 2)

# comparisons produce 1 for true and 0 for false

test(a < b, 1)
test(a <= b, 1)
test(a >= b, 0)
test(a > b, 0)
test(a == b, 0)
test(a != b, 1)

# if_else() can be used instead of branching
# let's find out the larger number
test((a < b).if_else(b, a), 2)

# arrays and loops work as follows

a = Array(100, sint)

@for_range(100)
def f(i):
    a[i] = sint(i) * sint(i - 1)

test(a[99], 99 * 98)

# if you use loops, use Array to store results
# don't do this
# @for_range(100)
# def f(i):
#     a = sint(i)
# test(a, 99)

# sfix: fixed-point numbers

# set the precision after the dot and in total

sfix.set_precision(16, 32)

# and the output precision in decimal digits

print_float_precision(4)

# you can do all basic arithmetic with sfix, including division

a = sfix(2)
b = sfix(-0.1)

test(a + b, 1.9)
test(a - b, 2.1)
test(a * b, -0.2)
test(a / b, -20)
test(a < b, 0)
test(a <= b, 0)
test(a >= b, 1)
test(a > b, 1)
test(a == b, 0)
test(a != b, 1)

test((a < b).if_else(a, b), -0.1)

# now let's do a computation with private inputs
# party 0 supplies three number and party 1 supplies three percentages
# we want to compute the weighted mean

print_ln('Party 0: please input three numbers not adding up to zero')
print_ln('Party 1: please input any three numbers')

data = Matrix(3, 2, sfix)

# use @for_range_opt for balanced optimization
# but use Python loops if compile-time numbers are need (e.g., for players)

@for_range_opt(3)
def _(i):
    for j in range(2):
        data[i][j] = sfix.get_input_from(j)

# compute weighted average

weight_total = sum(point[0] for point in data)
result = sum(point[0] * point[1] for point in data) / weight_total

# branching is supported also depending on revealed secret data
# with garbled circuits this triggers a interruption of the garbling

@if_e((sum(point[0] for point in data) != 0).reveal())
def _():
    print_ln('weighted average: %s', result.reveal())
@else_
def _():
    print_ln('your inputs made no sense')

# permutation matrix

M = Matrix(2, 2, sfix)
M[0][0] = 0
M[1][0] = 1
M[0][1] = 1
M[1][1] = 0

# matrix multiplication

M = data * M
test(M[0][0], data[0][1].reveal())
test(M[1][1], data[1][0].reveal())