feat(collect): 按本次构建生成 collect 的工具链 + 回环回归测试
tools/tdc_collectgen.py: 静态导出槽位序 -> 按 #36 规则算「项->槽」-> 重排 -> 该构建密钥加密。 同一构建实测: 37 项 -> 38 槽, 明文 686B / collect 920 字符, 解密回环逐字节一致。 规则修正(重要): * kind=1 的串只按**括号外**逗号切槽 —— '[]'/'[360,360]'/'["zh-CN","zh"]' 内逗号不构成新槽 * kind=2 的项**仍产出 1 槽**(r3 清空转存 r4, r4 在收尾 @4905 拼回) 新增 tcaptcha/tests/test_collectgen.py (已并入 selftest) 与 fixture: outer_commas / 槽数推算 == 真机槽数 / 重排是置换 / 加解密回环 / 不匹配必须报错。 账本新增已知 #37(工具链跑通) 与 #38(纯离线生成尚未打通, 诚实标注)。
This commit is contained in:
yml2213
@@ -66,7 +66,9 @@
|
||||
| 33 | **VM 程序结构层已吃透：顶层就是 62 个模块工厂的注册表** | 顶层 `@0` 做 `registry[id] = function(module, exports, require){...}`（`DUP; PUSH_IMM [id]; REVERSE_TOP2; JUMP -> CREATE_CLOSURE; SET_PROP`）。`tools/tdc_program.py --modules` 直接导出。修了两个真 bug：① 反汇编器漏了 `RET_TRUE`/`THROW` 是终结符（控制流穿过函数结尾 → 子程序归属全错）；② 提升器终结符判定写死 opcode 数值 | 确证 |
|
||||
| 34 | **`cd` 槽位序的来源已定位并可静态导出** | 装配链：清单模块 `arr[arr.length] = require(N)`（**追加，所以 require 顺序就是数组顺序**，已核）→ 分发模块按 `.on`/`.reset`/`.get` 分三数组 → 装配模块 `r3='{"cd":['` + `for(...) r3 += arr[i].get()`。线上构建实测 **37 个 require + 1 个字面量槽 = 38**，与真机 `cd` 槽数完全吻合；两份构建的 require 串不同（与 #31 一致）。工具：`tools/tdc_layout.py` | 确证（顺序）/ 语义标签部分 |
|
||||
| 35 | **运行时反证通过：静态导出的 require 序 == 运行时真实 require 序（逐项一致）** | 在真机 iframe（`tdc.js` 实际执行的上下文，`window.TDC.getData` 就在这里）里包一层 `Function.prototype.apply`，模块工厂调用形状 = **4 参 `[object,object,object,function]`**（require 是最后一个），按工厂实例分组记账。触发一次刷新后取运行时串，用 `tools/cdp_script_source.js`（Debugger.scriptParsed + getScriptSource）取到**运行时真正执行的那份 tdc.js**（exactly-once，不能用重新 fetch 代替——#29 每次返回不同构建），两边串：<br>静态 = 运行时 = `[8,9,10,13,14,15,16,17,18,19,20,21,23,24,25,30,31,32,34,36,37,1,38,39,40,41,42,43,44,47,48,49,50,51,56,57,58]`（37 项，逐位相同）<br>→ 由此可判定：之前"部分槽语义对不上"是**标注启发式**的问题，**顺序本身正确**。工具：`scripts/experiment_require_order.py` | 确证 |
|
||||
| 36 | **「清单项 → cd 槽」的精确规则（同一构建端到端校验确立）** | 装配循环（线上构建 `@4362`，370 条）分支树已读通：<br>`if(!item) continue;` `v,w,kind = item.get();`<br>· **`kind===2`（`@4820`）→ 不进 `cd`**：它把 `r3` 清空（`r3=""`）转存到 `r4`（`@4860`），`r4` 在收尾 `@4905` 被二次加工后另作他用（如随机串 `vXc29RxhvT34Q…`）<br>· `v===null/undefined`（`@4799`）→ `out += "null"`<br>· `typeof v==="number"`（`@4738`）→ `out += (v||0)`<br>· **`kind===1`（`@4724`）→ `out += v` 原样拼**：返回值是 JSON 碎片串，串里的逗号会**切出额外槽**（实测 `"641547489,1791732001"` → 2 槽）<br>· 其余（`@4703`）→ `out += '"' + v + '"'`<br>· 收尾 `@4753`：`typeof w === "number"` 才再 +1 槽（实测 kind=2 的第二元素是 null，不触发；此路径本次未命中）<br>→ **37 项 → 38 槽**，仅第 33 项（模块 1）占 2 槽。工具：`tools/tdc_slotmap.py`（`--items` 运行时取样 + `--plain` 明文对拍） | 确证 |
|
||||
| 36 | **「清单项 → cd 槽」的精确规则（同一构建端到端校验确立）** | 装配循环（线上构建 `@4362`，370 条）分支树已读通：<br>`if(!item) continue;` `v,w,kind = item.get();`<br>· **`kind===2`（`@4820`）**：把 `r3` 清空转存 `r4`（`@4860`），`r4` 在收尾 `@4905` 二次加工后**拼回**，因此**仍产出 1 槽**（实测第 0 项，随机串 `vXc29RxhvT34Q…`）<br>· `v===null/undefined`（`@4799`）→ `out += "null"`<br>· `typeof v==="number"`（`@4738`）→ `out += (v||0)`<br>· **`kind===1`（`@4724`）→ `out += v` 原样拼**：返回值是 JSON 碎片串；**只按括号外的逗号**切槽 —— `'[]'`/`'[360,360]'`/`'["zh-CN","zh"]'` 内的逗号不构成新槽，而 `"641547489,1791732001"` → 2 槽<br>· 其余（`@4703`）→ `out += '"' + v + '"'`<br>· 收尾 `@4753`：`typeof w === "number"` 才再 +1 槽（实测 kind=2 的第二元素是 null，不触发；此路径本次未命中）<br>→ **37 项 → 38 槽**，仅第 33 项（模块 1）占 2 槽。工具：`tools/tdc_slotmap.py`（`--items` 运行时取样 + `--plain` 明文对拍） | 确证 |
|
||||
| 37 | **按构建生成 collect 的工具链已跑通（含回环自测）** | `tools/tdc_collectgen.py`：静态导出槽位序 → 按 #36 规则算「项→槽」→ 重排 → 该构建密钥加密。<br>实测（同一构建）：37 项 → **38 槽**，明文 686B / collect 920 字符；**解密回环逐字节一致**，cd 与真机样本逐槽相同。<br>规则修正（重要）：`kind=1` 的串**只按括号外逗号**切槽 —— `'[]'`/`'[360,360]'`/`'["zh-CN","zh"]'` 内的逗号不构成新槽；`kind=2` 的项**仍产出 1 槽**（转存 r4 后在收尾 @4905 拼回）。<br>回归：`tcaptcha/tests/test_collectgen.py`（已并入 selftest） | 确证 |
|
||||
| 38 | **纯离线生成还没打通（诚实标注）** | 当前 `--from-sample` 依赖**一份同机的真机明文**做重排；`--values`（语义名→值）需要调用方提供取值字典。**仍缺**：纯算地为每个语义名算出值（多数是静态环境量可配置，但 canvas/WebGL 指纹只能取自真实环境）。这一步完成前，不能声称"纯离线过风控"。 | 状态说明 |
|
||||
|
||||
---
|
||||
|
||||
|
||||
@@ -16,6 +16,7 @@ SUITES = [
|
||||
"tcaptcha.tests.test_glyph_solver",
|
||||
"tcaptcha.tests.test_icon_click_solver",
|
||||
"tcaptcha.tests.test_slider_solver",
|
||||
"tcaptcha.tests.test_collectgen",
|
||||
"tcaptcha.tests.test_xtea_key",
|
||||
"tcaptcha.tests.test_tdc_key",
|
||||
]
|
||||
|
||||
@@ -0,0 +1 @@
|
||||
{"seq": [8, 9, 10, 15, 18, 19, 20, 21, 27, 28, 31, 33, 34, 36, 37, 38, 39, 40, 41, 26, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 55, 1, 56, 57, 58], "kinds": [2, 1, null, null, null, null, null, null, 1, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, 1, null, null, null, null, null, 1, null, 1, null], "v0s": ["vXc29RxhvT34Qjg4BO81Rcg6kcS36LwS", "1", "1440", "0", "0", "0", "2560", "98k", "[]", "+08", "https://turing.captcha.gtimg.com/1/template/drag_ele.html?rand=1519713624347", "1440-2560-2530-24-*-*-|-*", "0", "MacIntel", "2", "https://1ms.run/console/personal/benefits/check-in", "1791736469", "", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/155.0.0.0 Safari/537.36", "1984571394", "ANGLE (Apple, ANGLE Metal Renderer: Apple M4, Unspecified Version)", "iframe", "", "0", "1023", "Google Inc. (Apple)", "", "[360,360]", "1791736491", "24", "UTF-8", "10", "GgoAAAANSUhEUgAAASwAAACWCAYAAATf3v8bh+ryWcfSBSLkmN3N1PQinPL7Bi3TG0qE2CogAAAABJRU5ErkJggg==", "641547489,1791732001", "2", "[\"zh-CN\",\"zh\"]", "0"], "sample_cd": [1791736322, 2560, [[1, 1, 12]], 0, "1440-2560-2530-24-*-*-|-*", 1440, "iframe", 0, "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/155.0.0.0 Safari/537.36", "ANGLE (Apple, ANGLE Metal Renderer: Apple M4, Unspecified Version)", "https://turing.captcha.gtimg.com/1/template/drag_ele.html?rand=1519713624347", "", 0, "+08", 1023, "", [360, 360], "UTF-8", "", 0, 2, 1791736260, 2086701610, 3, 1, "https://1ms.run/console/personal/benefits/check-in", "98k", [], "MacIntel", "Google Inc. (Apple)", "GgoAAAANSUhEUgAAASwAAACWCAYAAATf3v8bh+ryWcfSBSLkmN3N1PQinPL7Bi3TG0qE2CogAAAABJRU5ErkJggg==", 0, 10, 641547489, 1791732001, 24, ["zh-CN", "zh"], 0]}
|
||||
@@ -0,0 +1,105 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""
|
||||
按构建生成 collect 的回归测试（承 PROGRESS.md #34/#35/#36；AGENTS.md §5.9）。
|
||||
|
||||
固化的性质:
|
||||
1. `outer_commas`: 只数**括号外**的逗号 —— 方括号内的逗号不构成顶层分隔
|
||||
（'[]' / '[360,360]' / '["zh-CN","zh"]' 拼回去仍是一个值）；
|
||||
2. `SlotSpec` 推算出的槽数 == 真机明文 cd 的实际槽数；
|
||||
3. 重排是**置换**：值集合不变（排序后逐项相同）；
|
||||
4. 加密/解密**回环一致**（明文逐字节还原）；
|
||||
5. 槽数不匹配的样本**必须报错**，不能静默错位。
|
||||
|
||||
样本: tcaptcha/tests/fixtures/tdc_slotmap_case.json
|
||||
= {seq, kinds, v0s, sample_cd}，取自 2026-10-11 真机会话（37 项 -> 38 槽）。
|
||||
|
||||
运行:
|
||||
.venv/bin/python -m tcaptcha.tests.test_collectgen
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import json
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[2]
|
||||
if str(ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(ROOT))
|
||||
if str(ROOT / "tools") not in sys.path:
|
||||
sys.path.insert(0, str(ROOT / "tools"))
|
||||
|
||||
from tcaptcha.crypto import xtea_decrypt_bytes # noqa: E402
|
||||
|
||||
import tdc_collectgen as cg # noqa: E402
|
||||
|
||||
FIX = ROOT / "tcaptcha" / "tests" / "fixtures" / "tdc_slotmap_case.json"
|
||||
# 固定密钥，仅用于回环自洽（不代表任何真实构建）
|
||||
KEY = [1497711441, 1213874789, 1799903316, 1835232858]
|
||||
|
||||
_fails: list = []
|
||||
|
||||
|
||||
def check(name: str, cond: bool, detail: str = "") -> None:
|
||||
print(f" [{'PASS' if cond else 'FAIL'}] {name}" + (f" {detail}" if detail else ""))
|
||||
if not cond:
|
||||
_fails.append(name)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
if not FIX.exists():
|
||||
print(f"缺少 fixture: {FIX}")
|
||||
return 1
|
||||
case = json.loads(FIX.read_text(encoding="utf-8"))
|
||||
seq, kinds, v0s, sample_cd = case["seq"], case["kinds"], case["v0s"], case["sample_cd"]
|
||||
print(f"fixture: 项 {len(seq)} / kind {len(kinds)} / v0 {len(v0s)} / 真机 cd {len(sample_cd)}")
|
||||
|
||||
print("\n[1] outer_commas: 只数括号外的逗号")
|
||||
check("'[]' -> 0", cg.outer_commas("[]") == 0)
|
||||
check("'[360,360]' -> 0", cg.outer_commas("[360,360]") == 0)
|
||||
check("'[\"zh-CN\",\"zh\"]' -> 0", cg.outer_commas('["zh-CN","zh"]') == 0)
|
||||
check("'641547489,1791732001' -> 1", cg.outer_commas("641547489,1791732001") == 1)
|
||||
check("'a,b,c' -> 2", cg.outer_commas("a,b,c") == 2)
|
||||
|
||||
print("\n[2] 槽数推算 == 真机明文槽数")
|
||||
spec = cg.SlotSpec(seq, kinds, v0s)
|
||||
print(f" {spec.describe()}")
|
||||
check("槽数一致", spec.n_slots == len(sample_cd),
|
||||
f"{spec.n_slots} vs {len(sample_cd)}")
|
||||
check("fixture 覆盖 kind=1", any(k == 1 for k in kinds))
|
||||
check("fixture 覆盖 kind=2", any(k == 2 for k in kinds))
|
||||
|
||||
print("\n[3] 重排是置换（值集合不变）")
|
||||
out = cg.reorder(sample_cd, spec)
|
||||
key = lambda v: json.dumps(v, sort_keys=True, ensure_ascii=False)
|
||||
check("长度不变", len(out) == len(sample_cd), f"{len(out)} vs {len(sample_cd)}")
|
||||
check("值集合相同（排序后逐项相等）",
|
||||
sorted(map(key, out)) == sorted(map(key, sample_cd)))
|
||||
|
||||
print("\n[4] 加密/解密回环")
|
||||
plain = cg.build_plain(out)
|
||||
b64 = cg.encrypt_plain(plain, KEY)
|
||||
back = xtea_decrypt_bytes(base64.b64decode(b64), KEY).decode("latin1").rstrip("\x00").strip()
|
||||
check("明文逐字节还原", back == plain)
|
||||
back_obj = json.loads(back)
|
||||
check("cd 还原", back_obj["cd"] == out)
|
||||
check("sd 还原", back_obj["sd"] == {"od": "C", "ft": "qf_7Pf__H"})
|
||||
|
||||
print("\n[5] 槽数不匹配必须报错")
|
||||
raised = False
|
||||
try:
|
||||
cg.reorder(sample_cd[:-1], spec)
|
||||
except ValueError:
|
||||
raised = True
|
||||
check("少一槽 -> ValueError", raised)
|
||||
|
||||
print(f"\n{'=' * 62}")
|
||||
if _fails:
|
||||
print(f"❌ collectgen 测试失败: {_fails}")
|
||||
return 1
|
||||
print("✅ collectgen 测试全部通过")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,218 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""
|
||||
tdc_collectgen.py —— **按本次构建**生成 collect 明文（并加密）。
|
||||
|
||||
承 har/1ms_dianxuan/PROGRESS.md #29~#36:
|
||||
* #29/#31 槽位序**按构建随机化**，不能用缓存/旧模板；
|
||||
* #34/#35 槽位序 = 清单模块的 require 顺序，**可纯静态导出**；
|
||||
* #36 「清单项 → cd 槽」不是一一对应：`kind===2` 的项不进 cd，
|
||||
`kind===1` 的返回值是 JSON 碎片串、串内逗号会切出额外槽。
|
||||
|
||||
本模块做两件事:
|
||||
1. `reorder(template_cd, mapping, values)` —— 把"语义值"按**本次构建的槽位序**排成 cd；
|
||||
2. `build(...)` —— 组装成 `{"cd":[...],"sd":{...}}` 明文，并用该构建的 XTEA 密钥加密。
|
||||
|
||||
⚠️ 关于「语义值」从哪来（诚实说明，别把它当成已经解决的事）:
|
||||
目前有两种取值来源，都可用，但**完备性不同**：
|
||||
* `--values <json>`：由调用方提供「语义名 -> 值」字典（最稳，推荐）；
|
||||
* `--from-sample <cd json>`：用**另一份已解密的真机明文**做"同源重排"。
|
||||
这要求两份样本跑在**同一台机器/同一环境**（取值集合相同，只有顺序不同，见 #31），
|
||||
适用于"先拿一份真机明文当素材、再按本次构建重排"的验证场景。
|
||||
**仍需补齐**：纯算地把每个语义名对应的值算出来（多数是静态环境量，可配置；
|
||||
canvas/WebGL 指纹这类必须取自真实环境）。这一步未完成前，纯离线生成还不能保证过风控。
|
||||
|
||||
用法:
|
||||
# 用一份同机真机明文做重排, 按本次构建输出 collect
|
||||
.venv/bin/python tools/tdc_collectgen.py <tdc.js> \\
|
||||
--from-sample /tmp/real_cd.json --out /tmp/collect.b64
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import base64
|
||||
import json
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, List, Optional, Sequence, Tuple
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
if str(ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(ROOT))
|
||||
if str(ROOT / "tools") not in sys.path:
|
||||
sys.path.insert(0, str(ROOT / "tools"))
|
||||
|
||||
from tcaptcha.crypto import xtea_encrypt_bytes # noqa: E402
|
||||
from tcaptcha.tdc_key import extract_xtea_key_pure # noqa: E402
|
||||
import tdc_layout # noqa: E402
|
||||
import tdc_program as tp # noqa: E402
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------- #
|
||||
# 槽位映射 #
|
||||
# --------------------------------------------------------------------------- #
|
||||
def outer_commas(s: str) -> int:
|
||||
"""数**不在**任何 `[]` / `{}` 内的逗号个数（方括号内的逗号不构成顶层分隔）。"""
|
||||
depth = 0
|
||||
n = 0
|
||||
for ch in s:
|
||||
if ch in "[{":
|
||||
depth += 1
|
||||
elif ch in "]}":
|
||||
depth = max(0, depth - 1)
|
||||
elif ch == "," and depth == 0:
|
||||
n += 1
|
||||
return n
|
||||
|
||||
|
||||
class SlotSpec:
|
||||
"""一次构建下的槽位规格（规则见 PROGRESS.md #36）。"""
|
||||
|
||||
def __init__(self, seq: Sequence[int],
|
||||
kinds: Optional[Sequence[Any]] = None,
|
||||
v0s: Optional[Sequence[str]] = None) -> None:
|
||||
self.seq = list(seq)
|
||||
self.kinds = list(kinds) if kinds is not None else [None] * len(seq)
|
||||
# v0s[i] = 该项 get()[0] 的字符串化结果；kind=1 时按它的逗号数切出额外槽
|
||||
self.v0s = list(v0s) if v0s is not None else [""] * len(seq)
|
||||
self.assign: List[Tuple[int, int, int, int]] = [] # (item, module, first_slot, n_slots)
|
||||
self._build()
|
||||
|
||||
def _build(self) -> None:
|
||||
slot = 0
|
||||
for i, m in enumerate(self.seq):
|
||||
k = self.kinds[i]
|
||||
n = 1
|
||||
if k == 1 and isinstance(self.v0s[i], str):
|
||||
# 原样拼 -> 该串会被 JSON 解析器按逗号切成多个值。
|
||||
# 但**数组/对象形态**的串（'[]' / '[360,360]' / '["zh-CN","zh"]'）里的逗号
|
||||
# 处在方括号内, 拼回去仍是一个值, 不产生新槽 —— 只数"括号外"的逗号。
|
||||
n += outer_commas(self.v0s[i])
|
||||
if k == 2:
|
||||
# kind=2 把 r3 清空转存 r4（不接在 cd 后面），但 r4 会在收尾 @4905
|
||||
# 被二次加工后重新拼进 r3 —— 实测它**仍然产出一个槽**。
|
||||
n = 1
|
||||
self.assign.append((i, m, slot, n))
|
||||
slot += n
|
||||
|
||||
@property
|
||||
def n_slots(self) -> int:
|
||||
return sum(a[3] for a in self.assign)
|
||||
|
||||
def describe(self) -> str:
|
||||
multi = [(i, m, n) for (i, m, _s, n) in self.assign if n > 1]
|
||||
return (f"项 {len(self.seq)} -> 槽 {self.n_slots} "
|
||||
f"(kind=2 跳过 {sum(1 for k in self.kinds if k == 2)} 项"
|
||||
+ (f"; 多槽项 {multi}" if multi else "") + ")")
|
||||
|
||||
|
||||
def load_shape(items_path: str, seq: Sequence[int]) -> Tuple[List[Any], List[str]]:
|
||||
"""从运行时取样(items json)里读出每项的 kind 与 get()[0] 的字符串化值。"""
|
||||
d = json.load(open(items_path, encoding="utf-8"))
|
||||
res = d.get("result", d)
|
||||
rows = res.get("out") or res.get("items") or []
|
||||
kinds: List[Any] = []
|
||||
v0s: List[str] = []
|
||||
for r in rows:
|
||||
if not isinstance(r, dict):
|
||||
kinds.append(None); v0s.append(""); continue
|
||||
kinds.append(r.get("kind"))
|
||||
v = r.get("v0")
|
||||
v0s.append(str(v) if v is not None else "")
|
||||
if len(kinds) < len(seq):
|
||||
pad = len(seq) - len(kinds)
|
||||
kinds += [None] * pad
|
||||
v0s += [""] * pad
|
||||
return kinds[: len(seq)], v0s[: len(seq)]
|
||||
|
||||
|
||||
def load_sample_cd(path: str) -> List[Any]:
|
||||
raw = json.load(open(path, encoding="utf-8"))
|
||||
return raw["cd"] if isinstance(raw, dict) else raw
|
||||
|
||||
|
||||
def reorder(sample: Sequence[Any], spec: SlotSpec) -> List[Any]:
|
||||
"""把"上一份 cd"的值按**本次构建**的槽位序重排。
|
||||
|
||||
原理（#31）：同一台机器上的多份样本，**取值集合相同**、只有顺序不同；
|
||||
所以只要样本与本次构建的"槽位数一致"，就能按"项 -> 槽"的对应关系做置换。
|
||||
"""
|
||||
if len(sample) != spec.n_slots:
|
||||
raise ValueError(
|
||||
f"样本槽数 {len(sample)} != 本次构建槽数 {spec.n_slots}；"
|
||||
f"两者必须来自同一环境（见 PROGRESS.md #31）")
|
||||
out: List[Any] = []
|
||||
for (_i, _m, slot, n) in spec.assign:
|
||||
out.extend(sample[slot:slot + n])
|
||||
return out
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------- #
|
||||
# 明文组装 + 加密 #
|
||||
# --------------------------------------------------------------------------- #
|
||||
def build_plain(cd: Sequence[Any], sd: Optional[dict] = None) -> str:
|
||||
sd = sd or {"od": "C", "ft": "qf_7Pf__H"}
|
||||
return json.dumps({"cd": list(cd), "sd": sd}, separators=(",", ":"))
|
||||
|
||||
|
||||
def encrypt_plain(plain: str, key: Sequence[int]) -> str:
|
||||
data = plain.encode("latin1")
|
||||
data += b"\x00" * ((8 - len(data) % 8) % 8) # 与真实样本一致: \x00 补齐
|
||||
return base64.b64encode(xtea_encrypt_bytes(data, list(key))).decode("ascii")
|
||||
|
||||
|
||||
def main() -> int:
|
||||
ap = argparse.ArgumentParser(description="按本次构建生成 collect")
|
||||
ap.add_argument("path", help="本次会话的 tdc.js")
|
||||
ap.add_argument("--from-sample", default=None, help="一份已解密的 cd 数组(json)")
|
||||
ap.add_argument("--values", default=None, help="语义名->值的 json 字典(实验性)")
|
||||
ap.add_argument("--out", default=None, help="输出 collect(base64) 的文件")
|
||||
ap.add_argument("--plain-out", default=None, help="同时输出明文")
|
||||
ap.add_argument("--kinds", default=None,
|
||||
help="运行时取样的 items json（提供每项的 kind，见 #36）")
|
||||
args = ap.parse_args()
|
||||
|
||||
prog = tp.load(args.path)
|
||||
_mid, _entry, seq = tdc_layout.find_manifest_module(prog)
|
||||
key = extract_xtea_key_pure(open(args.path, encoding="utf-8", errors="replace").read())["key"]
|
||||
|
||||
kinds, v0s = load_shape(args.kinds, seq) if args.kinds else (None, None)
|
||||
spec = SlotSpec(seq, kinds, v0s)
|
||||
print(f"文件 : {args.path}")
|
||||
print(f"清单模块 : registry[{_mid}] 项数 {len(seq)}")
|
||||
print(f"槽位规格 : {spec.describe()}")
|
||||
print(f"XTEA 密钥 : {key}")
|
||||
|
||||
if args.from_sample:
|
||||
sample = load_sample_cd(args.from_sample)
|
||||
try:
|
||||
cd = reorder(sample, spec)
|
||||
except ValueError as exc:
|
||||
print(f"❌ {exc}")
|
||||
return 1
|
||||
print(f"样本槽数 : {len(sample)} -> 重排后 {len(cd)}")
|
||||
elif args.values:
|
||||
vals = json.load(open(args.values, encoding="utf-8"))
|
||||
cd = []
|
||||
for (_i, m, _slot, n) in spec.assign:
|
||||
v = vals[str(m)]
|
||||
cd.extend(v[:n] if isinstance(v, list) else [v])
|
||||
print(f"按语义字典 : {len(cd)} 槽")
|
||||
else:
|
||||
print("需要 --from-sample 或 --values")
|
||||
return 1
|
||||
|
||||
plain = build_plain(cd)
|
||||
b64 = encrypt_plain(plain, key)
|
||||
print(f"明文长度 : {len(plain)}")
|
||||
print(f"collect : {len(b64)} 字符 (tlg={len(b64)})")
|
||||
if args.plain_out:
|
||||
Path(args.plain_out).write_text(plain, encoding="latin1")
|
||||
print(f"明文 -> {args.plain_out}")
|
||||
if args.out:
|
||||
Path(args.out).write_text(b64, encoding="ascii")
|
||||
print(f"collect -> {args.out}")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Reference in New Issue
Block a user