Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Python agentic-aiai-securityalert-triageblue-teamcybersecuritydetection-engineeringincident-responselanggraphllm-agentsmcpmitre-attackmodel-context-protocolpurple-teamsecurity-operationsself-hostedsiemsoarsocthreat-detectionthreat-intelligence
"""Round-trip tests for the ChatOps callback HMAC tokens.
This module is security-critical: the only thing protecting the callback
endpoint from forged "the user said yes" responses is the signature on
these tokens. Every check below is here because if it silently regresses,
a malicious chat user could acknowledge prompts that were never sent to
them, suppressing legitimate alerts.
"""
from __future__ import annotations
import time
from uuid import uuid4
import pytest
from app.security.chatops_token import (
ChatOpsTokenError,
mint_token,
verify_token,
)
_SECRET = "test-secret-do-not-use-in-prod"
def _mint(**overrides):
"""Mint a token with sensible defaults so individual tests can override
only the field they care about."""
base = {
"action_id": uuid4(),
"case_id": uuid4(),
"tenant_id": uuid4(),
"choice": "acknowledge",
"user_ref": "[email protected]",
"secret": _SECRET,
"ttl_seconds": 900,
}
base.update(overrides)
return base, mint_token(**base)
def test_round_trip_returns_original_claims():
args, token = _mint()
claims = verify_token(token, _SECRET)
assert claims.action_id == args["action_id"]
assert claims.case_id == args["case_id"]
assert claims.tenant_id == args["tenant_id"]
assert claims.choice == "acknowledge"
assert claims.user_ref == "[email protected]"
assert claims.expires_at > claims.issued_at
def test_tampered_payload_fails_signature():
_, token = _mint()
payload, sig = token.split(".", 1)
# Flip a single character in the payload — should invalidate HMAC.
tampered_payload = payload[:-1] + ("A" if payload[-1] != "A" else "B")
with pytest.raises(ChatOpsTokenError, match="invalid_signature"):
verify_token(f"{tampered_payload}.{sig}", _SECRET)
def test_wrong_secret_fails_signature():
_, token = _mint()
with pytest.raises(ChatOpsTokenError, match="invalid_signature"):
verify_token(token, "different-secret")
def test_expired_token_is_rejected():
_, token = _mint(ttl_seconds=1)
# Verify in the future, past the TTL.
with pytest.raises(ChatOpsTokenError, match="expired"):
verify_token(token, _SECRET, now=int(time.time()) + 5)
def test_unsupported_choice_is_rejected_at_mint_time():
"""Tokens for unknown choices never get minted, so the callback handler
never has to make a policy call about how to log them."""
with pytest.raises(ChatOpsTokenError, match="unsupported choice"):
mint_token(
action_id=uuid4(),
case_id=uuid4(),
tenant_id=uuid4(),
choice="confirm-everything", # not in the allowed set
user_ref="alice",
secret=_SECRET,
ttl_seconds=900,
)
def test_missing_secret_at_mint_or_verify_is_rejected():
with pytest.raises(ChatOpsTokenError):
mint_token(
action_id=uuid4(),
case_id=uuid4(),
tenant_id=uuid4(),
choice="deny",
user_ref="alice",
secret="",
ttl_seconds=900,
)
_, token = _mint()
with pytest.raises(ChatOpsTokenError):
verify_token(token, "")
def test_malformed_token_fails_clean():
with pytest.raises(ChatOpsTokenError, match="malformed"):
verify_token("not-a-valid-token", _SECRET)
with pytest.raises(ChatOpsTokenError, match="malformed"):
verify_token("a.b.c", _SECRET)
def test_each_choice_produces_distinct_tokens():
"""Defensive: same action+case but different choice must produce
different signatures. Otherwise the callback can't tell which button
the user pressed."""
base = {
"action_id": uuid4(),
"case_id": uuid4(),
"tenant_id": uuid4(),
"user_ref": "alice",
"secret": _SECRET,
"ttl_seconds": 900,
}
ack = mint_token(choice="acknowledge", **base)
deny = mint_token(choice="deny", **base)
escalate = mint_token(choice="escalate", **base)
assert len({ack, deny, escalate}) == 3