好的，好是主观的，但它更适合标题。我真正想知道的是，这种密码管理/身份验证策略是否存在明显的缺陷。我也对性能很好奇。我正在尝试使用 PBKDF2 进行身份验证，但我不确定这对于 RESTful Web 服务是否是个好主意。我可以在这里和那里找到这个问题的某些部分答案，但我从未找到一个全面的自上而下的答案。好了，这就是我的全部策略。
背景：
- Service 是 C# 中的 ASP.NET Web Api 项目
- 后端是 MS SQL 服务器
- 凭证使用基本身份验证通过 HTTPS 发送，即“授权：基本用户名：密码”
- API 将被 Andriod 应用、ios 应用和网站使用
一、后端：
create table [dbo].[User] (
[Name] varchar(50) collate SQL_Latin1_General_CP1_CI_AI not NULL primary key clustered,
[Password] varchar(28) collate SQL_Latin1_General_CP1_CS_AS not NULL,
[Salt] varchar(28) not NULL)
创建用户的存储过程：
create procedure [dbo].[User_Create]
@Name varchar(50),
@Salt varchar(50),
@Password varchar(50)
as
insert into [dbo].[User]([Name], [Salt], [Password])
values(@Name, @Salt, @Password)
以及获取用户的存储过程：
create procedure [dbo].[User_Get]
@Name varchar(50)
as
select *
from [dbo].[User]
where [Name] = @Name
对于后端，我很好奇为名称、密码和盐选择的数据类型是否合适。
这是创建新用户并将其持久保存到后端的代码。这可能是我最关心的，无论是从安全角度还是从性能角度。
public void CreateUser(string username, string password)
{
int hashLength = 20;
int saltLength = 20;
int hashIterations = 1000;
using(SqlConnection connection = new SqlConnection(this._ConnectionString))
using (SqlCommand command = new SqlCommand("[dbo].[User_Create]"))
{
Rfc2898DeriveBytes pbkdf2 = new Rfc2898DeriveBytes(password, saltLength, hashIterations);
string salt = Convert.ToBase64String(pbkdf2.Salt);
string hashPassword = Convert.ToBase64String(pbkdf2.GetBytes(hashLength));
command.CommandType = CommandType.StoredProcedure;
command.Parameters.AddWithValue("@Name", username);
command.Parameters.AddWithValue("@Salt", salt);
command.Parameters.AddWithValue("@Password", hashPassword);
connection.Open();
command.ExecuteNonQuery();
}
}
这是验证用户的代码。这实际上只是执行存储过程，调用ValidatePassword执行 PBKDF2 并将User对象返回给设置 Principal 的 Web API 消息处理程序（角色尚未实现）：
public User AuthenticateUser(string username, string password)
{
using (SqlConnection connection = new SqlConnection(this._ConnectionString))
using (SqlCommand command = new SqlCommand("[dbo].[User_Get]"))
{
command.CommandType = CommandType.StoredProcedure;
command.Parameters.AddWithValue("@Name", username);
command.Connection = connection;
connection.Open();
using (SqlDataReader reader = command.ExecuteReader())
{
if (reader.Read() && this.ValidatePassword(password, reader["Salt"].ToString(), reader["Password"].ToString()))
{
return new User()
{
Name = username,
CustomerId = reader["CustomerId"].ToString()
};
}
else
{
return null;
}
}
}
}
这里是ValidatePassword，上面的代码依赖它来验证密码（以防不明显）。我还想确保我做对了。
private bool ValidatePassword(string password, string salt, string hashedPassword)
{
int hashLength = 20;
int hashIterations = 1000;
byte[] saltBytes = Convert.FromBase64String(salt);
Rfc2898DeriveBytes pbkdf2 = new Rfc2898DeriveBytes(password, saltBytes, hashIterations);
byte[] hashBytes = pbkdf2.GetBytes(hashLength);
string hash = Convert.ToBase64String(hashBytes);
// Security Decisions For String Comparisons
//
// If you are making a security decision (such as whether to allow access to a system resource) based on the
// result of a string comparison or a case change, you should not use the invariant culture. Instead, you
// should perform a case-sensitive or case-insensitive ordinal comparison by calling a method that includes
// a StringComparison parameter and supplying either StringComparison.Ordinal or
// StringComparison.OrdinalIgnoreCase as an argument. Code that performs culture-sensitive string operations
// can cause security vulnerabilities if the current culture is changed or if the culture on the computer
// that is running the code differs from the culture that is used to test the code. In contrast, an ordinal
// comparison depends solely on the binary value of the compared characters.
//
// Source: http://msdn.microsoft.com/en-us/library/system.globalization.cultureinfo.invariantculture.aspx
return hash.Equals(hashedPassword, StringComparison.Ordinal);
}