通过 chatgpt 协助完成网站数据破解

Chatgpt 的出现极大地提升了程序员的工作效率，常见的使用场景包括代码自动生成、代码静态检查等，那么 chatgpt 能否用于某些网站的数据破解工作呢？

问题

某天线上服务开始报警，原来是某个视频网站无法获取到其 cdn 地址导致的下载失败问题。

经过 debug 发现原来的明文数据现在变成了加密数据（数据放在 html 中），如下

由于职责所在，需要对此加密数据进行破解

解决方案

通过搜索，可以定位到相关 js 代码位置，点击 Open in Sources panel，可以进行断点调试

结合断点和变量输出，我们发现核心的解密函数逻辑是 Object(le.videoDataDecrypt)，点击可以继续跟踪

这下有点傻眼了，这都是什么啊

不要急，让我们翻译翻译

那么，解密逻辑

return e = a[t(540)][t(737)][t(577)](a[t(540)][t(506)][t(582)](e)), (0, o[t(782)])(e, t(747))

等价于

e = a['enc']['Utf8']['stringify'](a['enc']['Base64']['parse'](e));

return (0, o['xorCipher'])(e, 'guanhui456')

我们观察到，`a['enc']['Base64']['parse']` 看起来像是 base64 解码，经过尝试发现确实是；

而 `a['enc']['Utf8']['stringify']`，容易获得其源码为

stringify: function(e) {
    var t = l;
    try {
        return decodeURIComponent(escape(v[t(577)](e)))
    } catch (r) {
        throw new Error(t(754))
    }
}

所以，解密逻辑变成了

e = decodeURIComponent(escape(base64decode(e)))
return (0, o['xorCipher'])(e, 'guanhui456')

`(0, o['xorCipher'])` 的逻辑可以直接通过 chatgpt 进行理解和生成

结合上面这些分析，可以得到测试的 python 代码如下

import base64
from urllib.parse import unquote, quote

e = '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'


def simple_decode(data):
    # 先进行 base64 解码
    data = base64.b64decode(data)
    # 再进行 escape 和 decodeURIComponent 操作
    data = unquote(quote(data), encoding='utf-8')
    return data


def xor_strings(data, key):
    n = ""
    for i in range(len(data)):
        o = ord(data[i])
        s = ord(key[i % len(key)])
        n += chr(o ^ s)
    return n


print(xor_strings(simple_decode(e), 'guanghui456'))

总结

chatgpt 可以用在 javascript 破解的两个方面

代码理解：对于一些难以读懂的 javascript 代码，可以让 gpt 辅助进行理解。有些可能是已经成熟的算法，chatgpt 可以识别出来
代码改写：将 javascript 代码改写为其他语言，例如 python 或 java，便于集成进业务代码中

写在最后

实际第一次做的时候，完全没有文章中写的这么顺利，那个 base64decode 的代码，也是靠 gpt 猜出来的。。。