Yolov/model_crypto.py

# model_crypto.py
import os
import tempfile

import requests
from cryptography.fernet import Fernet
from cryptography.hazmat.primitives import hashes
from cryptography.hazmat.primitives.kdf.pbkdf2 import PBKDF2HMAC
import base64
import pickle
from pathlib import Path
from log import logger


class ModelEncryptor:
    """模型加密/解密器"""

    @staticmethod
    def generate_key(password: str, salt: bytes = None):
        """生成加密密钥"""
        if salt is None:
            salt = os.urandom(16)

        kdf = PBKDF2HMAC(
            algorithm=hashes.SHA256(),
            length=32,
            salt=salt,
            iterations=100000,
        )
        key = base64.urlsafe_b64encode(kdf.derive(password.encode()))
        return key, salt

    @staticmethod
    def encrypt_model(model_path: str, output_path: str, password: str):
        """加密模型文件"""
        try:
            # 读取模型文件
            with open(model_path, 'rb') as f:
                model_data = f.read()

            # 生成密钥
            key, salt = ModelEncryptor.generate_key(password)
            fernet = Fernet(key)

            # 加密数据
            encrypted_data = fernet.encrypt(model_data)

            # 保存加密数据（包含salt）
            encrypted_payload = {
                'salt': salt,
                'data': encrypted_data,
                'original_size': len(model_data)
            }

            with open(output_path, 'wb') as f:
                pickle.dump(encrypted_payload, f)

            logger.info(f"模型加密成功: {model_path} -> {output_path}")
            return True

        except Exception as e:
            logger.error(f"模型加密失败: {str(e)}")
            return False

    @staticmethod
    def decrypt_model(encrypted_path: str, password: str):
        """解密模型到内存"""
        try:
            # 读取加密文件
            with open(encrypted_path, 'rb') as f:
                encrypted_payload = pickle.load(f)

            salt = encrypted_payload['salt']
            encrypted_data = encrypted_payload['data']

            # 生成密钥
            key, _ = ModelEncryptor.generate_key(password, salt)
            fernet = Fernet(key)

            # 解密数据
            decrypted_data = fernet.decrypt(encrypted_data)

            # 保存到临时文件
            with tempfile.NamedTemporaryFile(delete=False, suffix='.pt') as tmp:
                tmp.write(decrypted_data)
                temp_path = tmp.name

            logger.info(f"模型解密成功: {encrypted_path}")
            return temp_path

        except Exception as e:
            logger.error(f"模型解密失败: {str(e)}")
            return None

    @staticmethod
    def is_encrypted(model_path: str):
        """检查模型是否加密"""
        try:
            with open(model_path, 'rb') as f:
                # 尝试读取加密格式
                data = pickle.load(f)
                return isinstance(data, dict) and 'salt' in data and 'data' in data
        except:
            return False


class ModelManager:
    """模型管理器，支持加密模型加载"""

    def __init__(self, config):
        self.config = config
        self.models_dir = "models"
        os.makedirs(self.models_dir, exist_ok=True)

    def load_model(self, model_config):
        """加载模型（支持加密）"""
        model_path = model_config['path']
        encrypted = model_config.get('encrypted', False)
        encryption_key = model_config.get('encryption_key')

        local_path = os.path.join(self.models_dir, os.path.basename(model_path))

        # 下载模型（如果不存在）
        if not os.path.exists(local_path):
            if not self.download_model(model_config):
                return None

        # 如果是加密模型，需要解密
        if encrypted and encryption_key:
            if ModelEncryptor.is_encrypted(local_path):
                decrypted_path = ModelEncryptor.decrypt_model(local_path, encryption_key)
                if decrypted_path:
                    try:
                        from ultralytics import YOLO
                        model = YOLO(decrypted_path).to(model_config['device'])

                        # 清理临时文件
                        try:
                            os.unlink(decrypted_path)
                        except:
                            pass

                        return model
                    except Exception as e:
                        logger.error(f"加载解密模型失败: {str(e)}")
                        return None
                else:
                    return None
            else:
                logger.warning(f"模型未加密或密钥错误: {local_path}")
                return None
        else:
            # 普通模型加载
            try:
                from ultralytics import YOLO
                model = YOLO(local_path).to(model_config['device'])

                # 应用配置
                if model_config.get('half', False) and 'cuda' in model_config['device']:
                    model = model.half()

                return model
            except Exception as e:
                logger.error(f"加载模型失败: {str(e)}")
                return None

    def download_model(self, model_config):
        """下载模型"""
        try:
            model_path = model_config['path']
            download_url = model_config.get('download_url')

            if not download_url:
                logger.error(f"模型无下载地址: {model_path}")
                return False

            local_path = os.path.join(self.models_dir, os.path.basename(model_path))

            logger.info(f"下载模型: {download_url} -> {local_path}")

            response = requests.get(download_url, stream=True, timeout=30)
            response.raise_for_status()

            total_size = int(response.headers.get('content-length', 0))
            downloaded = 0

            with open(local_path, 'wb') as f:
                for chunk in response.iter_content(chunk_size=8192):
                    downloaded += len(chunk)
                    f.write(chunk)

                    if total_size > 0:
                        progress = downloaded * 100 // total_size
                        if progress % 10 == 0:
                            logger.info(f"下载进度: {progress}%")

            logger.info(f"模型下载完成: {local_path} ({downloaded} 字节)")
            return True

        except Exception as e:
            logger.error(f"下载模型失败: {str(e)}")
            return False