feat(ttp): split bash CMD evidence into structured uid/user/src/pwd/cmd rows

The inspector was dumping the whole `CMD uid=0 user=root src=… pwd=…
cmd=nmap -p- 192.168.1.0/24` syslog body into a single ``command_text``
blob. ANTI: "I'd like to separate the fields." Done — three layers
work together:

1. Collector session aggregator: new `_parse_cmd_msg` splits the bash
   PROMPT_COMMAND msg into `{uid, user, src, pwd, command}`. The
   session-ended envelope's per-command dict now carries the
   structured fields, with `command_text` set to just the cmd= value
   (preserving embedded whitespace — `nmap -p- 1.2.3.0/24` etc.).

2. Rule engine: per-source_kind auxiliary evidence list
   (`_AUX_EVIDENCE_FIELDS`). For `command` events the engine
   automatically promotes uid/user/src/pwd into the persisted
   `evidence` dict on top of the rule's explicit `evidence_fields`.
   Engine-controlled, not per-rule — adding a new aux field is one
   line here, not a 30-rule YAML sweep, and rule authors can't
   accidentally drop it.

3. TTPInspector frontend: evidence renders as a structured
   `kvs` grid (UID / USER / SRC / PWD / CMD rows) instead of
   pretty-printed JSON. Primary-order list keeps shell fields at
   the top; everything else falls below alphabetically so unfamiliar
   evidence shapes still surface predictably.

Tests:
- session_aggregator pins the structured-fields emit (uid/user/src/
  pwd/command_text without "CMD" prefix, embedded whitespace
  preserved).
- rule_engine_tagger pins the aux-field auto-promotion + the
  no-`None`-leakage path when payload doesn't carry an aux key.
This commit is contained in:
2026-05-02 03:20:53 -04:00
parent 84699f89da
commit d1c4a48963
6 changed files with 268 additions and 4 deletions

View File

@@ -31,6 +31,25 @@ def _cmd(ts_iso: str, text: str) -> dict[str, Any]:
}
def _raw_cmd(ts_iso: str, msg: str) -> dict[str, Any]:
"""Parsed event whose bash CMD body is in ``msg``, fields={}.
Mirrors what the unmodified collector parser produces for
PROMPT_COMMAND lines (the parser deliberately keeps fields empty
so the frontend pill rendering doesn't double-up). The aggregator
now extracts uid/user/src/pwd/command from that msg body.
"""
return {
"timestamp": ts_iso,
"decky": "SRV-DELTA-77",
"service": "bash",
"event_type": "command",
"attacker_ip": _ATTACKER_IP,
"fields": {},
"msg": msg,
}
def _session_recorded(
ts_iso: str, sid: str, duration_s: float = 60.0,
) -> dict[str, Any]:
@@ -196,6 +215,36 @@ def test_ttl_eviction_drops_old_commands() -> None:
assert remaining == ["fresh"]
def test_session_emits_structured_uid_user_src_pwd_when_msg_carries_them(
aggregator: _SessionAggregator,
captured_publishes: list[tuple[str, dict[str, Any], str]],
) -> None:
"""The bash PROMPT_COMMAND msg body splits into structured fields.
Pins the "inspector wants UID/SRC/PWD/CMD on separate rows"
contract. Without this the inspector sees one big
``CMD uid=0 user=root src=… cmd=…`` string and operators have to
eyeball the cmd= portion out of the prefix garbage.
"""
aggregator.add_event(_raw_cmd(
"2026-05-02T06:22:48",
"CMD uid=0 user=root src=192.168.1.5 pwd=/root "
"cmd=nmap -p- 192.168.1.0/24",
))
aggregator.add_event(_session_recorded(
"2026-05-02T06:23:00", sid="sess-x", duration_s=120.0,
))
payload = captured_publishes[0][1]
cmd = payload["commands"][0]
assert cmd["uid"] == "0"
assert cmd["user"] == "root"
assert cmd["src"] == "192.168.1.5"
assert cmd["pwd"] == "/root"
# ``command_text`` is the cmd= value, NOT the full "CMD uid=…" line.
# nmap's command line carries spaces — we must preserve them.
assert cmd["command_text"] == "nmap -p- 192.168.1.0/24"
def test_publish_failure_is_swallowed() -> None:
"""A blowing-up publish must not propagate into the stream thread."""
def _bad(_t: str, _p: dict[str, Any], _e: str) -> None: