feat: 增加人脸识别功能

apps/users/models/user/__init__.py

@@ -1,10 +1,14 @@
 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
 #
+import base64
+import struct
 import uuid
 
+import math
 from django.conf import settings
 from django.contrib.auth.models import AbstractUser, UserManager as _UserManager
+from django.core.exceptions import ValidationError
 from django.db import models
 from django.shortcuts import reverse
 from django.utils import timezone
@@ -23,13 +27,15 @@ from ._json import JSONFilterMixin
 from ._role import RoleMixin, SystemRoleManager, OrgRoleManager
 from ._source import SourceMixin, Source
 from ._token import TokenMixin
+from ._face import FaceMixin
 
 logger = get_logger(__file__)
 __all__ = [
     "User",
     "UserPasswordHistory",
     "MFAMixin",
-    "AuthMixin"
+    "AuthMixin",
+    "FaceMixin"
 ]
 
 
@@ -48,6 +54,7 @@ class User(
     TokenMixin,
     RoleMixin,
     MFAMixin,
+    FaceMixin,
     LabeledMixin,
     JSONFilterMixin,
     AbstractUser,
@@ -133,6 +140,9 @@ class User(
     slack_id = models.CharField(
         null=True, default=None, max_length=128, verbose_name=_("Slack")
     )
+    face_vector = fields.EncryptTextField(
+        null=True, blank=True, max_length=2048, verbose_name=_("Face Vector")
+    )
     date_api_key_last_used = models.DateTimeField(
         null=True, blank=True, verbose_name=_("Date api key used")
     )

apps/users/models/user/_face.py

@@ -0,0 +1,56 @@
+import base64
+import struct
+
+import math
+from django.conf import settings
+from django.core.exceptions import ValidationError
+
+from common.utils import (
+    get_logger,
+)
+
+logger = get_logger(__file__)
+
+
+class FaceMixin:
+    face_vector = None
+
+    def get_face_vector(self) -> list[float]:
+        if not self.face_vector:
+            raise ValidationError("Face vector is not set.")
+        return self._decode_base64_vector(str(self.face_vector))
+
+    def check_face(self, code) -> bool:
+        distance = self.compare_euclidean_distance(code)
+        similarity = self.compare_cosine_similarity(code)
+
+        return distance < settings.FACE_RECOGNITION_DISTANCE_THRESHOLD \
+            and similarity > settings.FACE_RECOGNITION_COSINE_THRESHOLD
+
+    def compare_euclidean_distance(self, base64_vector: str) -> float:
+        target_vector = self._decode_base64_vector(base64_vector)
+        current_vector = self.get_face_vector()
+        return self._calculate_euclidean_distance(current_vector, target_vector)
+
+    def compare_cosine_similarity(self, base64_vector: str) -> float:
+        target_vector = self._decode_base64_vector(base64_vector)
+        current_vector = self.get_face_vector()
+        return self._calculate_cosine_similarity(current_vector, target_vector)
+
+    @staticmethod
+    def _decode_base64_vector(base64_vector: str) -> list[float]:
+        byte_data = base64.b64decode(base64_vector)
+        return list(struct.unpack('<128d', byte_data))
+
+    @staticmethod
+    def _calculate_euclidean_distance(vec1: list[float], vec2: list[float]) -> float:
+        return sum((x - y) ** 2 for x, y in zip(vec1, vec2)) ** 0.5
+
+    @staticmethod
+    def _calculate_cosine_similarity(vec1: list[float], vec2: list[float]) -> float:
+        dot_product = sum(x * y for x, y in zip(vec1, vec2))
+        magnitude_vec1 = math.sqrt(sum(x ** 2 for x in vec1))
+        magnitude_vec2 = math.sqrt(sum(y ** 2 for y in vec2))
+        if magnitude_vec1 == 0 or magnitude_vec2 == 0:
+            raise ValueError("Vector magnitude cannot be zero.")
+        return dot_product / (magnitude_vec1 * magnitude_vec2)