某站视频真实地址解析过程
本帖最后由 fjqisba 于 2018-1-3 13:00 编辑前不久看到一篇关于BibiBili视频地址解析的源码，正好有兴趣，就自己也研究了研究,记录一下。
1.以http://www.bilibili.com/video/av12535537/ 此视频为例用HttpWatch抓包，很容易就能看出视频的地址是从https://interface.bilibili.com/playurl?quality=4&player=1&cid=20627240&ts=1504780011&sign=f37df88434579a4b009a2bee708a2a71&qn=112里获取的
按照常例分析下参数:
cid很容易猜测出是视频的ID号,在网页源码里面也能找到
Quality和qn是视频的清晰度
ts则是时间戳,这个也不用多说
关键就是这个sign，在HttpWatch中是完全搜不到任何相关的信息的，但网页访问了http://static.hdslb.com/play.swf这个文件，那么sign算法很可能就在这个播放器里。
2.下载这个SWF文件后，载入AS3 Sorcerer，代码一大片不是很好找
那就保存到文本里,在文本中查找”视频地址”这个关键字。
很快就能定位到关键的函数LoadPreview和LoadCidVideo那么接着搜索LoadPreview和LoadCidVideo定位到函数部分,可以看到代码如下
那么关键算法就是在getSign和getSign_v2里面了
public function getSign(_arg_1:String):String
{
var _local_2:int;
var _local_4 = null;
var _local_7:int;
var _local_8:int;
var _local_6:int;
var _local_5:int;
var _local_3:int = ESP;
_local_2 = _local_3;
_local_3 = (_local_3 - 48);
_local_5 = 16;
_local_6 = (_local_2 - 37);
ESP = (_local_3 & -16);
_local_7 = CModule.mallocString(_arg_1);
_local_8 = _arg_1.length;
do
{
var _local_9:int = (L__2E_str2 - _local_5);
_local_9 = li8((_local_9 + 16)) /*FlasCC (Alchemy)*/ ;
_local_3 = (_local_3 - 16);
si32(_local_9, (_local_3 + 4)); //FlasCC (Alchemy)
si32(_local_6, _local_3); //FlasCC (Alchemy)
ESP = _local_3;
F_sprintf();
_local_3 = (_local_3 + 16);
_local_5 = (_local_5 + -1);
_local_6 = (_local_6 + 2);
} while (_local_5 != 0);
_local_3 = (_local_3 - 16);
_local_9 = (_local_2 - 4);
si32(_local_9, (_local_3 + 12)); //FlasCC (Alchemy)
_local_9 = (_local_2 - 37);
si32(_local_9, (_local_3 + 8)); //FlasCC (Alchemy)
si32(_local_8, (_local_3 + 4)); //FlasCC (Alchemy)
si32(_local_7, _local_3); //FlasCC (Alchemy)
ESP = _local_3;
F_get_sign();
_local_3 = (_local_3 + 16);
_local_5 = eax;
_local_9 = li32((_local_2 - 4)) /*FlasCC (Alchemy)*/ ;
_local_4 = CModule.readString(_local_5, _local_9);
if (_local_7 != 0)
{
_local_3 = (_local_3 - 16);
si32(_local_7, _local_3); //FlasCC (Alchemy)
ESP = _local_3;
F_idalloc();
_local_3 = (_local_3 + 16);
};
if (_local_5 != 0)
{
_local_3 = (_local_3 - 16);
si32(_local_5, _local_3); //FlasCC (Alchemy)
ESP = _local_3;
F_idalloc();
_local_3 = (_local_3 + 16);
};
var _local_10 = _local_4;
_local_3 = _local_2;
ESP = _local_3;
return (_local_10);
}
}//package com.bilibili.interfaces
public function getSign_v2(_arg_1:String, _arg_2:int):String
{
var _local_12:*;
var _local_3:int;
var _local_5 = null;
var _local_6:int;
var _local_11:int;
var _local_8:int;
var _local_10:int;
var _local_9:int;
var _local_4:int = ESP;
_local_3 = _local_4;
_local_4 = (_local_4 - 48);
_local_6 = _arg_2;
if (_local_6 >= 5)
{
_local_12 = _arg_1;
}
else
{
_local_8 = (L__2E_str2 + (_local_6 << 4));
_local_9 = 16;
_local_10 = (_local_3 - 37);
ESP = (_local_4 & -16);
_local_6 = CModule.mallocString(_arg_1);
_local_11 = _arg_1.length;
do
{
var _local_7:int = (_local_8 - _local_9);
_local_7 = li8((_local_7 + 16)) /*FlasCC (Alchemy)*/ ;
_local_4 = (_local_4 - 16);
si32(_local_7, (_local_4 + 4)); //FlasCC (Alchemy)
si32(_local_10, _local_4); //FlasCC (Alchemy)
ESP = _local_4;
F_sprintf();
_local_4 = (_local_4 + 16);
_local_9 = (_local_9 + -1);
_local_10 = (_local_10 + 2);
} while (_local_9 != 0);
_local_4 = (_local_4 - 16);
_local_7 = (_local_3 - 4);
si32(_local_7, (_local_4 + 12)); //FlasCC (Alchemy)
_local_7 = (_local_3 - 37);
si32(_local_7, (_local_4 + 8)); //FlasCC (Alchemy)
si32(_local_11, (_local_4 + 4)); //FlasCC (Alchemy)
si32(_local_6, _local_4); //FlasCC (Alchemy)
ESP = _local_4;
F_get_sign();
_local_4 = (_local_4 + 16);
_local_10 = eax;
_local_7 = li32((_local_3 - 4)) /*FlasCC (Alchemy)*/ ;
_local_5 = CModule.readString(_local_10, _local_7);
if (_local_6 != 0)
{
_local_4 = (_local_4 - 16);
si32(_local_6, _local_4); //FlasCC (Alchemy)
ESP = _local_4;
F_idalloc();
_local_4 = (_local_4 + 16);
};
if (_local_10 != 0)
{
_local_4 = (_local_4 - 16);
si32(_local_10, _local_4); //FlasCC (Alchemy)
ESP = _local_4;
F_idalloc();
_local_4 = (_local_4 + 16);
};
_local_12 = _local_5;
};
_local_4 = _local_3;
ESP = _local_4;
return (_local_12);
}
}//package com.bilibili.interfaces
可以看到，二者代码非常相似，而且光看代码无法找到有用的信息，那么就得想办法去调试了
3.目前我只知道的办法是：使用JPEXS反编译软件来修改SWF代码(插Log),然后通过Fiddler劫持替换SWF,通过显示Log信息来进行调试{:301_971:}
要显示log信息就得安装debug版本的flash
http://www.adobe.com/support/flashplayer/debug_downloads.html
为方便阅读日志信息可以安装Cygwin
具体步骤我也是参考了http://blog.csdn.net/hot_vc/article/details/50600717这篇文章{:301_988:}
————————————————————————————————————————————————————————————————————————————
将SWF载入JPEXS
要想修改SWF的源码就得去修改PCODE，第一次接触这种也不是很懂，觉得有点类似于C#的IL代码
通过AS3代码和PCODE代码的一一进行对比，还是能找到一些规律的
为了查看代码中变量值，用到以下代码
findpropstrict Qname(PackageNamespace(""),"trace")
getlocal 5
callpropvoid Qname(PackageNamespace(""),"trace") 1
这个代码就相当于trace(_loc5_);
修改保存后利用Fiddler来劫持替换我们的SWF
这样Cygwin就能输出我们想要看到的结果了。
其中有些变量的值是地址，而不是数据，为了读取地址中的数据，利用到了这句代码CModule.readString(地址,读取长度);
按照原PCODE的格式，写出插LOG时的PCODE
pushint 60
setlocal 8
getlex Qname(PackageNamespace("com.bilibili.interfaces"),"CModule")
getlocal 7
getlocal 8
callproperty Qname(PackageNamespace(""),"readString") 2
coerce_s
setlocal 5
findpropstrict Qname(PackageNamespace(""),"trace")
getlocal 5
callpropvoid Qname(PackageNamespace(""),"trace") 1
此代码相当于
_loc8_:int = 60;
_loc5_:* = CModule.readString(_loc7_,_loc8_);
trace(_loc5_);
4.其实关键算法在F_get_sign()中，因为从F_get_sign()函数中出来后，eax中就已经存放带着sign的视频URL的地址了。
F_get_sign()函数过于复杂,总之通过不断地插这两种Log查看变量信息,最终找到了算法{:301_1008:}
由这个地方可以看出是MD5,还有一个位置能够准确地查看到进行MD5处理的数据(已经忘了.....)
算法如下:
五种清晰度
qn=16,quality=1
qn=32,quality=5
qn=48,quality=2
qn=64,quality=2
qn=80,quality=3
qn=112,quality=4
三种类型的视频bili2,bangumi,movie
bili2:
time = F_clock_gettime();//相当于取一个时间随机数
sign ＝ MD5(“cid=” ＋ cid ＋ “&player=1&qn=112&quality=4&ts=” ＋ time ＋ “1c15888dc316e05a15fdd0a02ed6584f”)
url=“https://interface.bilibili.com/playurl?player=1&qn=112&cid=” ＋ cid ＋ “&quality=4&ts=” ＋ time ＋ “&sign=” ＋ sign
bangumi:
time = F_clock_gettime();//相当于取一个时间随机数
sign ＝ MD5(“cid=” ＋ cid ＋ “&module=bangumi” ＋ “&player=1” ＋ “&qn=112” ＋ “&quality=4” ＋ “&ts=” ＋ time ＋ “9b288147e5474dd2aa67085f716c560d”)
url=“https://bangumi.bilibili.com/player/web_api/playurl?cid=” ＋ cid ＋ “&player=1” ＋ “&module=bangumi” ＋ “&qn=112” ＋ “&quality=4&ts=” ＋ time ＋ “&sign=” ＋ sign
movie:
time = F_clock_gettime();//相当于取一个时间随机数
sign ＝ MD5(“cid=” ＋ cid ＋ “&module=movie” ＋ “&player=1” ＋ “&qn=112” ＋ “&quality=4” ＋ “&ts=” ＋ time ＋ “9b288147e5474dd2aa67085f716c560d”)
url=“https://bangumi.bilibili.com/player/web_api/playurl?cid=” ＋ cid ＋ “&player=1” ＋ “&module=movie” ＋ “&qn=112” ＋ “&quality=4&ts=” ＋ time ＋ “&sign=” ＋ sign
也就只分析到了这里{:301_999:}
顺便说下，有些tx.acgvideo.com开头的视频无法下载是因为服务器检查了Refer,只要在协议头里面加入Refer:https://static.hdslb.com/play.swf,就能下载了
水平有限，分析若有什么错误，大家看着办吧{:301_998:} 学习到了大牛解决问题的思路,点赞 膜拜一番 感谢楼主分享心得，学习了！ 太牛了，要是有个视频教程看更好 学习一下. 借来研究 一下…… 学习思路 谢分享 感谢分享 对了.我想知道谁有解析解析音频节目的思路