mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 15:31:59 +00:00
ThearchyHelios
b080302e77
上一轮让被改过、解析不出的口径卡不再原样嵌入（号码不进文件），可这样一来它在页面上和正常的遮罩卡一模一样：「? 无法核对 含遮罩列，未随文件附带」，总结论从「有 1 项不通过」变成「可核对的项全部通过」，篡改的证据丢了；截断的卡服务端其实认不出输入，却被说成「含遮罩列」。现在 _Artifacts.add 对只给 id 的遮罩工件（口径卡、查询快照都算，带遮罩的快照被改过原来就有同样的问题）照实带上服务端看到的：store.intact 为假标 mismatch，解析不出对象标 unparsable。verifier.js 在遮罩分支认这两个标记：mismatch 判不通过，原因「导出时服务端的这份文件与工件 id（哈希）不一致，可能被改过；按含遮罩列处理，未随文件附带」；unparsable 是正常的无法核对，原因「导出时服务端的这份文件无法解析，认不出是否含遮罩列，按含遮罩列处理，未随文件附带」，引用它的数字跟着用这句。pytest 补被改过、截断、不是对象、哈希对得上却解析不出的卡和被改过的遮罩快照；check-bundle 在 masked-card 样本上补两条篡改检查。
180 lines
11 KiB
Python
180 lines
11 KiB
Python
"""可验证报告里的口径卡和遮罩（期 4 I2a 复核）：口径卡上记着每个输入的值和代入式，有输入取自遮罩列时整张卡只给 id。
|
||
|
||
以前只有查询快照会因遮罩不随文件附带，口径卡一律原样嵌入：证据面板上已经遮住的号码（cell() 取的遮罩格、按位置取的
|
||
vars.res.rows[0][1]），在这份要发给别人的文件里照样看得到。认哪个输入取自哪份查询结果的哪几列，和证据面板、推演同一个
|
||
认法（recompute.input_origin）；核对不了的宁可多挡。
|
||
|
||
服务端的卡文件被改过（哈希对不上）也照卡上记的输入认：嵌进去的是原字节，号码照样在里面。读得出字节、却解析不出一张卡
|
||
（截断的 JSON、不是对象）的，认不出输入，整张按带遮罩算、不嵌：和读不出来的快照按带遮罩算是同一个道理。
|
||
|
||
不嵌的工件，服务端把它知道的情况照实带上：自己的文件和工件 id（哈希）对不上的标 mismatch，解析不出来的标 unparsable，
|
||
核对页据此判不通过、照实说原因（以前原字节嵌进去，浏览器自己算出对不上；现在不嵌了，这个证据不能跟着丢）。带遮罩的
|
||
查询快照同样标。
|
||
|
||
工件、事件都是手搭的（会员、号码都是编的），照 tests/test_bundle_build.py 的样子。
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import base64
|
||
|
||
import pytest
|
||
|
||
from app.bundle.build import build_bundle, render_html
|
||
from app.core.artifact_store import content_hash
|
||
from tests.test_bundle_build import GRAPH, MASKED_ID, RUN, SNAP_MASKED, STORE, doc_of, stored
|
||
|
||
#: SNAP_MASKED 里的号码（phone 设了遮罩）
|
||
PHONE = SNAP_MASKED["rows"][0][1]
|
||
CELL = "cell(nodes.members, 0, 'phone')"
|
||
|
||
|
||
def card_with(inputs, substituted="len('000-0001')", expression=f"len({CELL})"):
|
||
"""会员号码位数一张卡：报告引用的是它（m:gmv 指过来）。"""
|
||
return {"kind": "metric_set", "caliber": "会员口径", "caliber_version": "v1", "metrics": [
|
||
{"id": "gmv", "name": "号码位数", "value": 1280.5, "unit": "元", "decimals": 1, "expression": expression,
|
||
"substituted": substituted, "inputs": inputs}]}
|
||
|
||
|
||
def bundle_of(card, *, outputs=None, source_masks=None, tamper=None, rehash=False, files_changed=None):
|
||
"""一次手搭的运行：会员查询（带遮罩的快照）→ 数据整形（outputs 里的产出）→ 口径卡 → 报告。tamper 给了就把服务端的
|
||
卡文件换成 tamper(原字节)，id 照旧（文件被改过、哈希对不上）；rehash 为真时 id 按换过的字节算（服务端存的就是这一份，
|
||
哈希对得上）。files_changed 换掉服务端别的文件（{工件 id: 新字节}，id 照旧）。"""
|
||
card_id, card_raw = stored(card)
|
||
if tamper is not None:
|
||
card_raw = tamper(card_raw)
|
||
if rehash:
|
||
card_id = content_hash(card_raw)
|
||
doc = doc_of(extra_catalog={"m:gmv": {"alias": "m:gmv", "kind": "metric", "artifact": card_id, "node_id": "card",
|
||
"locator": {"metric": "gmv"}, "value": 1280.5}})
|
||
doc_id, doc_raw = stored(doc)
|
||
files = {**STORE, card_id: card_raw, doc_id: doc_raw, **(files_changed or {})}
|
||
events = [
|
||
(1, "run.started", None, {"nodes": 7}),
|
||
(2, "tool.end", "members", {"tool": "db_query__crm", "query_artifact": MASKED_ID}),
|
||
(3, "node.finished", "members", {"evidence": [{"kind": "query", "node_id": "members", "artifact": MASKED_ID}]}),
|
||
]
|
||
for node_id, content in (outputs or {}).items():
|
||
artifact, raw = stored(content) if content is not None else ("ee" * 32, None)
|
||
if raw is not None:
|
||
files[artifact] = raw
|
||
events.append((len(events) + 1, "node.finished", node_id, {"artifact": artifact}))
|
||
events += [
|
||
(len(events) + 1, "node.started", "card", {}),
|
||
(len(events) + 2, "node.finished", "card", {"evidence": [{"kind": "metric_set", "node_id": "card",
|
||
"artifact": card_id}]}),
|
||
(len(events) + 3, "report.checked", "write", {"doc_artifact": doc_id, "ok": True}),
|
||
(len(events) + 4, "run.finished", None, {"output": {}}),
|
||
]
|
||
bundle = build_bundle(RUN, events, manifest_seq=len(events), manifest_hash="0" * 64, graph=GRAPH,
|
||
source_masks=source_masks or {}, read=files.get, exported_at="2026-10-06T08:00:00+00:00")
|
||
entry = next(a for a in bundle["artifacts"] if a["id"] == card_id)
|
||
return bundle, entry
|
||
|
||
|
||
def cell_input(column, value):
|
||
return {"path": f"cell(nodes.members, 0, '{column}')", "value": value, "node_id": "members", "via": "tool_cell",
|
||
"artifact": MASKED_ID, "locator": {"row": 0, "column": column}, "status": "verified"}
|
||
|
||
|
||
def test_a_card_reading_a_masked_cell_ships_only_its_id():
|
||
bundle, entry = bundle_of(card_with([cell_input("phone", PHONE)]))
|
||
assert entry == {"id": entry["id"], "kind": "metric_set", "node_id": "card", "masked": True}
|
||
assert PHONE not in render_html(bundle)
|
||
|
||
|
||
def test_a_card_reading_only_unmasked_columns_is_embedded():
|
||
"""同一份快照里没遮的列（消费额）：卡上没有遮罩列的值，照常嵌入、照常核对。"""
|
||
card = card_with([cell_input("spend", 320.0)], substituted="320.0", expression="cell(nodes.members, 0, 'spend')")
|
||
_, entry = bundle_of(card)
|
||
assert "masked" not in entry and base64.b64decode(entry["b64"]) == stored(card)[1]
|
||
|
||
|
||
def test_a_mask_set_on_the_source_now_counts_too():
|
||
card = card_with([cell_input("spend", 320.0)], substituted="320.0", expression="cell(nodes.members, 0, 'spend')")
|
||
_, entry = bundle_of(card, source_masks={"crm": ["spend"]})
|
||
assert entry.get("masked") is True and "b64" not in entry
|
||
|
||
|
||
PARSED = {"columns": SNAP_MASKED["columns"], "rows": SNAP_MASKED["rows"]}
|
||
POSITIONAL = "vars.res.rows[0][1]"
|
||
|
||
|
||
@pytest.mark.parametrize("parsed, masked", [
|
||
# 数据整形原样解析的查询结果（带着快照的 id）：认得出取自 phone 列
|
||
({**PARSED, "artifact": MASKED_ID, "source": "crm"}, True),
|
||
# 写变量的节点的产出读不出来：核对不了，宁可多挡
|
||
(None, True),
|
||
# 结果没存快照、只带着数据源名：查询当时记下的遮罩无从核对，宁可多挡
|
||
({**PARSED, "source": "crm"}, True),
|
||
# 改过形状、既没带快照的 id 也没带数据源名：认不出出处，照常嵌入（和证据面板、推演一样，是已知的缺口）
|
||
(PARSED, False),
|
||
])
|
||
def test_a_card_reading_a_cell_by_position_is_judged_the_same_way(parsed, masked):
|
||
item = {"path": POSITIONAL, "value": PHONE, "node_id": "parse", "via": "transform", "field": "rows[0][1]",
|
||
"status": "ok"}
|
||
card = card_with([item], substituted=f"len('{PHONE}')", expression=f"len({POSITIONAL})")
|
||
bundle, entry = bundle_of(card, outputs={"parse": parsed})
|
||
assert (entry.get("masked") is True) is masked and ("b64" in entry) is not masked
|
||
if masked:
|
||
assert PHONE not in render_html(bundle)
|
||
else:
|
||
assert PHONE in base64.b64decode(entry["b64"]).decode()
|
||
|
||
|
||
def test_a_card_file_changed_on_the_server_is_still_judged_by_its_inputs():
|
||
"""服务端的卡文件被改过（这里改了口径版本，哈希对不上）：照样按卡上记的输入认遮罩，记着号码的只给 id。以前先看哈希、
|
||
对不上就当卡上没有输入，原字节照样嵌进页面，号码进了文件。不嵌了，就另标 mismatch：核对页判这一项不通过，不和
|
||
没被改过的遮罩卡一样说成正常的无法核对。"""
|
||
card = card_with([cell_input("phone", PHONE)])
|
||
bundle, entry = bundle_of(card, tamper=lambda raw: raw.replace(b'"caliber_version":"v1"', b'"caliber_version":"v9"'))
|
||
assert entry["id"] == stored(card)[0]
|
||
assert entry == {"id": entry["id"], "kind": "metric_set", "node_id": "card", "masked": True, "mismatch": True}
|
||
assert PHONE not in render_html(bundle)
|
||
|
||
|
||
def test_a_changed_card_without_masked_inputs_is_still_embedded_for_the_browser_to_judge():
|
||
"""改过的卡上没有遮罩列的值：照常原样嵌进去，浏览器自己算出哈希对不上（只是遮罩的判断不再看哈希）。"""
|
||
card = card_with([cell_input("spend", 320.0)], substituted="320.0", expression="cell(nodes.members, 0, 'spend')")
|
||
changed = stored(card)[1].replace(b'"caliber_version":"v1"', b'"caliber_version":"v9"')
|
||
_, entry = bundle_of(card, tamper=lambda raw: changed)
|
||
assert "masked" not in entry and base64.b64decode(entry["b64"]) == changed
|
||
|
||
|
||
@pytest.mark.parametrize("damage", [
|
||
# 截断的 JSON：号码还在已经写下的那半截里
|
||
lambda raw: raw[: raw.index(b"000-0001") + len(b"000-0001")],
|
||
# 解析得出、却不是一张卡（对象）
|
||
lambda raw: b'["000-0001"]',
|
||
])
|
||
def test_a_card_file_that_does_not_parse_as_a_card_is_not_embedded(damage):
|
||
"""服务端的卡文件读得出字节、解析不出一张卡：认不出卡上有哪些输入，整张按带遮罩算、只给 id，不原样嵌进页面。字节和
|
||
id 对不上（mismatch），也解析不出来（unparsable）：两样都照实标上。"""
|
||
card = card_with([cell_input("phone", PHONE)])
|
||
bundle, entry = bundle_of(card, tamper=damage)
|
||
assert entry == {"id": entry["id"], "kind": "metric_set", "node_id": "card", "masked": True, "mismatch": True,
|
||
"unparsable": True}
|
||
assert PHONE not in render_html(bundle)
|
||
|
||
|
||
def test_a_card_stored_as_something_unparsable_is_marked_so():
|
||
"""服务端存的这一份本来就解析不出一张卡（哈希对得上，不是被改过）：整张按带遮罩算、不嵌，只标 unparsable——核对页照实
|
||
说认不出是否含遮罩列，不说成「含遮罩列」，也不说被改过。"""
|
||
card = card_with([cell_input("phone", PHONE)])
|
||
bundle, entry = bundle_of(card, tamper=lambda raw: raw[: raw.index(b"000-0001") + len(b"000-0001")], rehash=True)
|
||
assert entry == {"id": entry["id"], "kind": "metric_set", "node_id": "card", "masked": True, "unparsable": True}
|
||
assert PHONE not in render_html(bundle)
|
||
|
||
|
||
def test_a_masked_snapshot_changed_on_the_server_is_marked_so():
|
||
"""带遮罩的查询快照在服务端被改过（哈希对不上）：照旧只给 id，另标 mismatch，核对页判不通过。以前和没被改过的遮罩快照
|
||
一模一样，总结论还是「可核对的项全部通过」。没被改过的照旧只标遮罩。"""
|
||
card = card_with([cell_input("spend", 320.0)], substituted="320.0", expression="cell(nodes.members, 0, 'spend')")
|
||
snapshot = lambda bundle: next(a for a in bundle["artifacts"] if a["id"] == MASKED_ID) # noqa: E731
|
||
bundle, _ = bundle_of(card)
|
||
assert snapshot(bundle) == {"id": MASKED_ID, "kind": "query", "node_id": "members", "masked": True}
|
||
changed = stored(SNAP_MASKED)[1].replace("林一".encode(), "林二".encode())
|
||
bundle, _ = bundle_of(card, files_changed={MASKED_ID: changed})
|
||
assert snapshot(bundle) == {"id": MASKED_ID, "kind": "query", "node_id": "members", "masked": True,
|
||
"mismatch": True}
|
||
assert PHONE not in render_html(bundle)
|