从dotnet源码看文件上传绕waf

警告
本文最后更新于 2022-02-25,文中内容可能已过时。

# 前言

看了赛博群的《从commons-fileupload源码看文件上传绕waf》,文末提到了dotnet也有这种问题,于是看了下dotnet的源码。

# 环境

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
public ActionResult Index()
{
    if (Request.Files.Count>0)
    {
        var file = Request.Files[0];
        var filename = file.FileName;
        var contenttype = file.ContentType;
        var reader = new StreamReader(file.InputStream);
        var content = reader.ReadToEnd();
        var filepath = Request.MapPath("~/ ") + filename;
        file.SaveAs(filepath);
        var resp = $" filename:{filename}\n save file path:{filepath}\n file content:{content}\n file content type:{contenttype}";
        return Content(resp);
    }
    else
    {
        return Content("no file");
    }
}

# 分析

对于上传的文件处理类位于System.Web.HttpMultipartContentTemplateParser.Parse()函数

1
2
3
4
5
6
internal static MultipartContentElement[] Parse(HttpRawUploadedContent data, int length, byte[] boundary, Encoding encoding)
{
	HttpMultipartContentTemplateParser httpMultipartContentTemplateParser = new HttpMultipartContentTemplateParser(data, length, boundary, encoding);
	httpMultipartContentTemplateParser.ParseIntoElementList();
	return (MultipartContentElement[])httpMultipartContentTemplateParser._elements.ToArray(typeof(MultipartContentElement));
}

和Request.Files的层级调用关系如图

image.png

在FillInFilesCollection()中,content-type必须以multipart/form-data开头

image.png

这里和common fileupload的处理不同,然后进入this.GetMultipartContent()

image.png

1来处理boundary 2来解析上传文件流 主要看1

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
		private byte[] GetMultipartBoundary()
		{
			string text = HttpRequest.GetAttributeFromHeader(this.ContentType, "boundary");
			if (text == null)
			{
				return null;
			}
			text = "--" + text;
			return Encoding.ASCII.GetBytes(text.ToCharArray());
		}

GetAttributeFromHeader是关键函数

image.png

分号逗号和等于号作为分隔符,并根据字符集忽略一些空白字符

image.png

所以content-type可以这么写

1
Content-Type: multipart/form-data\u0085,;;,,,,,;;, boundary =   aaa

接着看GetMultipartContent函数,解析完boundary和文件内容流之后,进入3 Parse函数也就是我们开篇提到的函数。

image.png

Parse函数就直接跟进了ParseIntoElementList函数

image.png

其中1 ParsePartHeaders是关键函数

image.png

能看到这个函数用来解析Content-DispositionContent-Type,先以冒号分割拿到冒号后的部分

1
2
Content-Disposition: form-data; name="file"; filename="1.txt"
Content-Type: text/plain

form-data; name="file"; filename="1.txt"

再看ExtractValueFromContentDispositionHeader函数

image.png

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
string text = " " + name + "=";
int num = CultureInfo.InvariantCulture.CompareInfo.IndexOf(l, text, pos, CompareOptions.IgnoreCase);
if (num < 0)
{
	text = ";" + name + "=";
	num = CultureInfo.InvariantCulture.CompareInfo.IndexOf(l, text, pos, CompareOptions.IgnoreCase);
	if (num < 0)
	{
		text = name + "=";
		num = CultureInfo.InvariantCulture.CompareInfo.IndexOf(l, text, pos, CompareOptions.IgnoreCase);
	}
}

会自动加上分号和等号,所以可以随便构造

1
2
3
Content-Disposition:\u0;;;;!@#$%^&*(;asdas\u0085d;085filename=11.aspx
Content-Disposition:filename=11.aspx
Content-Disposition:aaaaaaaaaaafilename=11.aspx;aaaaaaaaa

两个对比一下就知道

1
Content-Disposition: form-data; name="file";filename="11.aspx"
  1. form-data字段可以不要
  2. 可以随便在filenamename前随意填充字段
  3. 但是filenamename后必须跟随等号,并且末尾有分号标识结束。

在ExtractValueFromContentDispositionHeader函数中会对取的值进行Trim()处理,也能用\u0085来处理

content-type同上

最后贴一张构造的图

image.png

# 写在文后

Request.Files[0]的name字段是忽略大小写的,Request.Files[0]和Request.Files[“file”]两种写法绕过时可能会出一些拿不到name的问题。

dotnet的特殊空白符如上文,但是位置一般只能放在两侧来用Trim去除。

文笔垃圾,措辞轻浮,内容浅显,操作生疏。不足之处欢迎大师傅们指点和纠正,感激不尽。