AST簡介
抽象語法樹(Abstract Syntax Tree,AST),或簡稱語法樹(Syntax tree),是源代碼語法結構的一種抽象表示。它以樹狀的形式表現編程語言的語法結構润樱,樹上的每個節(jié)點都表示源代碼中的一種結構。之所以說語法是“抽象”的羡棵,是因為這里的語法并不會表示出真實語法中出現的每個細節(jié)壹若。
使用到的第三方庫
esprima:代碼轉語法樹
estraverse:AST遍歷輔助庫(深度優(yōu)先搜索),包含遍歷及替換節(jié)點的功能
escodegen:語法樹轉代碼
jsdom:獲取模擬瀏覽器window對象(很多js會檢測環(huán)境)
相關文件
反混淆前后對比
-反混淆前
_0x1dc76b[_0x1972('0xed')](typeof window, _0x1dc76b[_0x1972('0x229')]) && _0x1dc76b[_0x1972('0x147')](typeof window[_0x1972('0x38e')], _0x1dc76b[_0x1972('0x229')]) ? _0x1dc76b[_0x1972('0x44c')](typeof window[_0x1972('0x38e')][_0x1972('0x131')], _0x1dc76b[_0x1972('0x299')]) ? /phantomjs/[_0x1972('0x348')](window[_0x1972('0x38e')][_0x1972('0x131')][_0x1972('0x39e')]()) ? ss = _0x1dc76b[_0x1972('0x64')](ss, _0x1dc76b[_0x1972('0x13')](0x1, 0xf)) : window[_0x1972('0x38e')][_0x1972('0x301')] ? ss = _0x1dc76b[_0x1972('0x64')](ss, _0x1dc76b[_0x1972('0x311')](0x1, 0xf)) : ss = _0x1dc76b[_0x1972('0x64')](ss, _0x1dc76b[_0x1972('0x36c')](0x1, 0x1)) : _0x1dc76b[_0x1972('0x44c')](typeof window[_0x1972('0x38e')][_0x1972('0x3bd')], _0x1dc76b[_0x1972('0x21f')]) ? ss = _0x1dc76b[_0x1972('0x64')](ss, _0x1dc76b[_0x1972('0x36c')](0x1, 0xf)) : !window[_0x1972('0x38e')][_0x1972('0x3bd')] ? ss = _0x1dc76b[_0x1972('0x64')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0xf)) : ss = _0x1dc76b[_0x1972('0x28c')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0xf)) : ss = _0x1dc76b[_0x1972('0x27c')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0xf));
_0x1dc76b[_0x1972('0x44c')](typeof window, _0x1dc76b[_0x1972('0x229')]) && _0x1dc76b[_0x1972('0x26')](typeof window[_0x1972('0x433')], _0x1dc76b[_0x1972('0x229')]) ? _0x1dc76b[_0x1972('0x20e')](_0x1dc76b[_0x1972('0x1d8')](window[_0x1972('0x433')][_0x1972('0x1da')], window[_0x1972('0x433')][_0x1972('0x176')]), 0x0) ? ss = _0x1dc76b[_0x1972('0x27c')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0x2)) : ss = _0x1dc76b[_0x1972('0x370')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0x10)) : ss = _0x1dc76b[_0x1972('0x34d')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0x10));
_0x1dc76b[_0x1972('0x26')](typeof document, _0x1dc76b[_0x1972('0x229')]) && _0x1dc76b[_0x1972('0x26')](typeof document[_0x1972('0x208')], fff) ? !document[_0x1972('0x208')](_0x1dc76b[_0x1972('0x454')]) ? ss = _0x1dc76b[_0x1972('0x2c6')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0x3)) : ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x323')](0x1, 0x11)) : ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x1a1')](0x1, 0x11));
_0x1dc76b[_0x1972('0x119')](typeof window, _0x1dc76b[_0x1972('0x229')]) ? _0x1dc76b[_0x1972('0x119')](typeof window[_0x1972('0x410')], _0x1dc76b[_0x1972('0x21f')]) ? ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x1a1')](0x1, 0x4)) : ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x18d')](0x1, 0x12)) : _0x1dc76b[_0x1972('0x32e')](typeof window, _0x1dc76b[_0x1972('0x229')]) && _0x1dc76b[_0x1972('0x363')](typeof window[_0x1972('0x197')], _0x1dc76b[_0x1972('0x21f')]) ? ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x18d')](0x1, 0x12)) : ss = _0x1dc76b[_0x1972('0x3da')](ss, _0x1dc76b[_0x1972('0x18d')](0x1, 0x12));
反混淆后
typeof window == _0x1dc76b['WfJBW'] && typeof window['navigator'] == _0x1dc76b['WfJBW'] ? typeof window['navigator']['userAgent'] == _0x1dc76b['XaIwC'] ? /phantomjs/['test'](window['navigator']['userAgent']['toLowerCase']()) ? ss = ss | 1 << 15 : window['navigator']['webdriver'] ? ss = ss | 1 << 15 : ss = ss | 1 << 1 : typeof window['navigator']['appVersion'] == _0x1dc76b['DWvnG'] ? ss = ss | 1 << 15 : !window['navigator']['appVersion'] ? ss = ss | 1 << 15 : ss = ss | 1 << 15 : ss = ss | 1 << 15;
typeof window == _0x1dc76b['WfJBW'] && typeof window['screen'] == _0x1dc76b['WfJBW'] ? 970 + 1680 > 0 ? ss = ss | 1 << 2 : ss = ss | 1 << 16 : ss = ss | 1 << 16;
typeof document == _0x1dc76b['WfJBW'] && typeof document['getElementById'] == fff ? !document['getElementById'](_0x1dc76b['djBAH']) ? ss = ss | 1 << 3 : ss = ss | 1 << 17 : ss = ss | 1 << 17;
typeof window == _0x1dc76b['WfJBW'] ? typeof window['callPhantom'] == _0x1dc76b['DWvnG'] ? ss = ss | 1 << 4 : ss = ss | 1 << 18 : typeof window == _0x1dc76b['WfJBW'] && typeof window['_phantom'] == _0x1dc76b['DWvnG'] ? ss = ss | 1 << 18 : ss = ss | 1 << 18;
說明
//類似這種調用方法皂冰,根據傳入參數算真實值(做了反混淆店展,會處理)
_0x1972('0x229')
//運算符封裝為方法的,都會進行反混淆
_0x4d51e8['GNVJM'] = function (_0x399108, _0x2aefef) {
return _0x399108 == _0x2aefef;
};
//某個key等于字符串秃流,(這種沒處理赂蕴,因為不是很影響讀碼了,就沒處理舶胀。)
_0x4d51e8['WfJBW'] = 'object';
逆向過程
1.下載核心js文件概说,內容混淆嚴重。先做反混淆嚣伐。分析代碼糖赔,拆解為幾種混淆規(guī)則,針對每種規(guī)則反混淆轩端。這個文件梳理主要有3放典、4種混淆規(guī)則,這里做了主要的2種,已可以閱讀七七八八了(主要就是幾個混淆奋构,然后又互相嵌套壳影,AST遍歷是深度優(yōu)先遍歷,會從內到外依次反混淆弥臼,可以處理嵌套)宴咧。
- 通過某個方法運算得到的string
var _0x1972 = function (_0x129385, _0x5b4f79) {
// ...............
//.............某些計算規(guī)則,返回根據傳入參數計算出來的string醋火。
//.................
return s;
}
//多次重復出現類似的代碼
_0x1972('0x49')
//=================================================================================================================================
//反混淆核心代碼
var ast = esprima.parseScript(content.toString());
//_0x1972方法在前57行悠汽,這個通過eval函數加載這個方法
eval(content.split('\n').slice(0, 57).join(''));
ast = estraverse.replace(ast, {
enter: function (node, parent) {
if (node.type === "CallExpression") {
//因為混淆文件會變化,變量名會變芥驳,函數名稱長度為7的只有_0x1972這個方法
if (node.callee.name && node.callee.name.length === 7 && node.arguments.length === 1) {
//如果是這個函數,則直接用計算后的結果替換
let data = eval(`${node.callee.name}('${node.arguments[0].value}')`);
if (data && data !== 'undefined' && data !== 'Number' && data !== 'NaN') {
return {
type: "Literal",
value: data,
raw: `'${data}'`
};
}
}
}
},
});
//=================================================================================================================================
//反混淆后_0x1972('0x49')變?yōu)橛嬎愫蟮闹礧fJBW
WfJBW
- 運算符混淆
_0x615b4b['uUzso'] = function (_0x59febf, _0x3b43de) {
return _0x59febf & _0x3b43de;
};
//中間有一次重命名過程
var _0x2bb501 = _0x615b4b;
//多次出現的類似代碼
_0x2bb501['uUzso'](_0xd72c68, 8)
//反混淆核心代碼
ast = estraverse.replace(ast, {
enter: function (node, parent) {
if (node.type === "CallExpression"
&& node.arguments.length === 2
&& node.callee
&& node.callee.property
&& node.callee.property.value
) {
let operator;
let rname;
let rvalue = node.callee.property.value;
//尋找重命名前的真實名稱
estraverse.traverse(ast, {
enter: function (find_node, parent) {
try {
if (find_node.type === "VariableDeclaration"
&& find_node.declarations[0].id.name === node.callee.object.name
) {
rname = find_node.declarations[0].init.name;
this.break()
}
} catch (e) {
this.skip()
}
}
});
//尋找對應方法的運算符茬高,賦值給operator
estraverse.traverse(ast, {
enter: function (find_node, parent) {
try {
if (find_node.type === "AssignmentExpression"
&& find_node.left.object.name === rname
&& find_node.left.property.value === rvalue
) {
operator = find_node.right.body.body[0].argument.operator;
if (operator) {
this.break()
}
this.skip()
}
} catch (e) {
this.skip()
}
}
});
//替換為簡化的語法樹
if (operator) {
node = {
"type": "BinaryExpression",
"operator": operator,
"left": node.arguments[0],
"right": node.arguments[1]
};
return node
}
}
}
}
);
//反混淆后代碼
_0x615b4b['uUzso'] = function (_0x59febf, _0x3b43de) {
return _0x59febf & _0x3b43de;
};
var _0x2bb501 = _0x615b4b;
//_0x2bb501['uUzso'](_0xd72c68, 8) 變?yōu)? _0xd72c68 & 8
_0xd72c68 & 8;
2兆旬、代碼反混淆后已可以閱讀,找到核心代碼怎栽。
//加密入口函數
ABC['prototype']['z'](seed, ts)
//之后會用python調用丽猬,這里簡單封裝入口
function encrypt(seed, ts) {
ts = Number(ts);
return ABC['prototype']['z'](seed, ts);
}
//hook取兩個真實參數,真實結果應為971bzdrUsH6rIWHBVAmYXWgD3LfYvX/ci0ZvZl1DAdpLwgm8rTU6n4iFR5UFL3dliWHPzK1oNJkfepDbUYXKr4ZeMTBkmYrZcjuFh8K6Z6LiCgJwowN9mPggIR+336cmJH9B
encrypt('pA2EJB+wiUR5EXY0ZDTsTSh2z4oxxDfBIGHbOzpdB6A=', 1589961615404);
3、nodejs環(huán)境熏瞄,添加jsdom模擬瀏覽器環(huán)境,返回的結果總變脚祟。而在瀏覽器環(huán)境下,結果不變强饮,且與之前hook取到的結果一致由桌。判斷反混淆未影響代碼邏輯,但是代碼中有判斷瀏覽器環(huán)境
//添加window對象
const {JSDOM} = jsdom;
const {window} = new JSDOM(`...`)
//有一些判斷環(huán)境的代碼(這里給列出部分)
//如 global邮丰、process行您、child_process這些都是nodejs環(huán)境有,而瀏覽器沒有剪廉。有50處左右對比環(huán)境的娃循。
if (typeof global != _0x3e1ab5['VMFpl'] || typeof process != _0x3e1ab5['VMFpl'] || typeof child_process != _0x3e1ab5['VMFpl'] || typeof Buffer != _0x3e1ab5['VMFpl'] || typeof sessionStorage == _0x3e1ab5['VMFpl']) {
ss = ss | 1 << 12;
}
4、根據上面找出的對比瀏覽器環(huán)境的代碼斗蒋,寫出對應的處理代碼如:
//導入jsdom捌斧,有部分檢測window對象,及其屬性
const jsdom = require("jsdom");
const {JSDOM} = jsdom;
//檢測當前url
const {window} = new JSDOM(`...`,{
url: "https://www.zhipin.com/job_detail/?query=",
});
// // or even
const {document} = (new JSDOM(`...`)).window;
//檢測top
top = window
//檢測setInterval方法toString的值泉沾,瀏覽器及nodejs環(huán)境有差異
setInterval['toString']= function () {
return "function setInterval() { [native code] }"
}
//有檢測部分差異變量
global = undefined
process = undefined
child_process = undefined
Buffer = undefined
sessionStorage = {}
5捞蚂、還有部分不好在代碼上覆蓋,直接在反混淆后替換
//這里是調試時過無限debugger
code = code.replace('debugger;', '')
//有一個eval函數內部也是無限debugger爆哑,這里用console.log替換eval洞难,直接輸出代碼不執(zhí)行
.replace('eval', 'console.log')
//替換部分瀏覽器值(在4中代碼最上方覆蓋不生效,這里直接替換掉,待研究)
.replace("window['screen']['availHeight']", '970')
.replace("window['screen']['availWidth']", '1680')
.replace("window['outerWidth']", '1680')
.replace("window['innerWidth']", '480')
.replace("window['outerHeight']", '967')
.replace("window['innerHeight']", '856')
6队贱、終于在nodejs環(huán)境執(zhí)行后與瀏覽器環(huán)境一致色冀,也與hook到的值一致。
7柱嫌、這個混淆的js文件锋恬,會定期更新,各變量名及內部加密算法改變编丘。這里的4和5的處理都加入反混淆腳本与学,在js文件改變時,必須反混淆后才能出現5中的代碼嘉抓。(1.因為這里是直接調用索守,所以不關心加密算法是什么。2.因為上面5的部分代碼無法覆蓋抑片,所以新的文件還是要走反混淆這一步卵佛。如果5的部分可以類似于4的處理,那么之后新js文件可以不用反混淆敞斋,只需要將4部分放入js文件頭即可)
后記
看似很復雜的混淆截汪,其實都只是幾種混淆規(guī)則、互相嵌套造成的植捎⊙媒猓看著腦殼疼,其實只要分析一下焰枢,借助AST蚓峦,完成部分反混淆規(guī)則,即可正常閱讀代碼医咨。