Python wrapper for labeled and unlabeled asymmetric private set intersection (APSI).
For manylinux_2_31_x86_64
compatible platforms you can install PyAPSI
from
PyPi with
pip install apsi
You can check the system library versions that are required to be
manylinux_2_31_x86_64
compatible in the
auditwheel policy.
NOTE: While AVX2 supported is currently patched out
(#11), APSI and its dependencies still seem
to choose optimizations during build time depending on the available CPU flags, which
can cause incompatibility of the pre-built wheels on older CPUs beyond what auditwheel
can identify (#13).
In case you feel like contributing a build setup for Windows and OSX compatible wheels or extend the "From Source" section below, I would be happy to review your pull request.
Example usage of the labeled APSI server and client. The unlabeled variant can be used analogous to this.
from apsi import LabeledServer, LabeledClient
apsi_params = """
{
"table_params": {
"hash_func_count": 3,
"table_size": 512,
"max_items_per_bin": 92
},
"item_params": {"felts_per_item": 8},
"query_params": {
"ps_low_degree": 0,
"query_powers": [1, 3, 4, 5, 8, 14, 20, 26, 32, 38, 41, 42, 43, 45, 46]
},
"seal_params": {
"plain_modulus": 40961,
"poly_modulus_degree": 4096,
"coeff_modulus_bits": [40, 32, 32]
}
}
"""
server = LabeledServer()
server.init_db(apsi_params, max_label_length=10)
server.add_items([("item", "1234567890"), ("abc", "123"), ("other", "my label")])
client = LabeledClient(apsi_params)
oprf_request = client.oprf_request(["item", "abc"])
oprf_response = server.handle_oprf_request(oprf_request)
query = client.build_query(oprf_response)
response = server.handle_query(query)
result = client.extract_result(response)
assert result == {"item": "1234567890", "abc": "123"}
To control multi threading and logging in APSI
see
apsi.utils
.
Before you start, make sure that Taskfile, Docker and Poetry are installed.
You can then run a full build with tests that will generate a wheel file in dist/
as
follows:
task wheel PYTHON_VERSION=3.10.4
Note: Only Python 3.8, 3.9, 3.10, and their patch versions for which official Python Docker images exist are supported.
Please have a look at the files inside
docker/
for the required vcpkg
setup and apsi
AVX2 patch, in case you'd like to build from source in a custom
environment.