Aegis/backend/app/services/atomic_import_service.py

"""Atomic Red Team import service.

Downloads the Atomic Red Team repository ZIP from GitHub, parses every
``atomics/T*/T*.yaml`` file, and upserts :class:`TestTemplate` records
into the database.

Strategy
--------
The GitHub REST API without authentication only allows 60 req/hour.
Since the Atomic Red Team repo contains 1 500+ YAML files we avoid
per-file requests entirely.  Instead we:

1. Download the full repo as a ZIP archive (~40 MB).
2. Extract in a temporary directory.
3. Walk ``atomics/T*/T*.yaml`` files parsing them with PyYAML.
4. Create / update ``TestTemplate`` rows keyed by ``atomic_test_id``.
5. Clean up the temporary directory.

Idempotency
-----------
Running the import twice does **not** create duplicates.  Existing
templates are identified by their ``atomic_test_id`` and simply skipped.
"""

import io
import logging
import os
import shutil
import tempfile
import zipfile
from pathlib import Path

import requests as _requests
import yaml
from sqlalchemy.orm import Session

from app.models.test_template import TestTemplate
from app.services.audit_service import log_action

logger = logging.getLogger(__name__)

# ---------------------------------------------------------------------------
# Constants
# ---------------------------------------------------------------------------

ATOMIC_RT_ZIP_URL = (
    "https://github.com/redcanaryco/atomic-red-team"
    "/archive/refs/heads/master.zip"
)

# Request timeout for the ZIP download (seconds)
_DOWNLOAD_TIMEOUT = 300

# Top-level directory name inside the ZIP
_ZIP_ROOT_PREFIX = "atomic-red-team-master"


# ---------------------------------------------------------------------------
# Internal helpers
# ---------------------------------------------------------------------------


def _download_zip(url: str = ATOMIC_RT_ZIP_URL) -> bytes:
    """Download the Atomic Red Team ZIP and return its raw bytes."""
    logger.info("Downloading Atomic Red Team ZIP from %s …", url)
    resp = _requests.get(url, timeout=_DOWNLOAD_TIMEOUT, stream=True)
    resp.raise_for_status()
    content = resp.content
    logger.info("Downloaded %.1f MB", len(content) / (1024 * 1024))
    return content


def _safe_extract_zip(zip_bytes: bytes, dest: str) -> None:
    """Extract *zip_bytes* into *dest* with Zip Slip and Zip Bomb protection.

    Raises :class:`ValueError` if any member tries to escape the target
    directory (path traversal / Zip Slip) or if the archive exceeds the
    safety limits.
    """
    # Maximum uncompressed size: 500 MB — prevents zip-bomb DoS
    _MAX_UNCOMPRESSED_SIZE = 500 * 1024 * 1024
    # Maximum number of entries
    _MAX_ENTRIES = 50_000

    dest_path = Path(dest).resolve()

    with zipfile.ZipFile(io.BytesIO(zip_bytes)) as zf:
        entries = zf.infolist()

        if len(entries) > _MAX_ENTRIES:
            raise ValueError(
                f"ZIP archive contains {len(entries)} entries "
                f"(limit: {_MAX_ENTRIES}) — possible zip bomb"
            )

        total_size = sum(info.file_size for info in entries)
        if total_size > _MAX_UNCOMPRESSED_SIZE:
            raise ValueError(
                f"ZIP uncompressed size {total_size / (1024 * 1024):.0f} MB "
                f"exceeds limit of {_MAX_UNCOMPRESSED_SIZE / (1024 * 1024):.0f} MB"
            )

        for member in entries:
            target = (dest_path / member.filename).resolve()
            if not target.is_relative_to(dest_path):
                raise ValueError(
                    f"Zip Slip detected — member '{member.filename}' "
                    f"resolves outside target directory"
                )

        zf.extractall(dest)


def _extract_zip(zip_bytes: bytes, dest: str) -> Path:
    """Extract *zip_bytes* into *dest* and return the path to the atomics/ dir."""
    _safe_extract_zip(zip_bytes, dest)
    atomics_dir = Path(dest) / _ZIP_ROOT_PREFIX / "atomics"
    if not atomics_dir.is_dir():
        raise FileNotFoundError(
            f"Expected atomics directory not found at {atomics_dir}"
        )
    return atomics_dir


def _parse_yaml_files(atomics_dir: Path) -> list[dict]:
    """Walk the atomics directory and parse all technique YAML files.

    Returns a flat list of dicts, each representing a single atomic test
    with the following keys::

        technique_id, index, name, description, platforms,
        executor_type, command, source_url
    """
    results: list[dict] = []
    yaml_files = sorted(atomics_dir.glob("T*/T*.yaml"))
    logger.info("Found %d YAML files to parse", len(yaml_files))

    for yaml_path in yaml_files:
        technique_id = yaml_path.stem  # e.g. "T1059.001"
        try:
            with open(yaml_path, "r", encoding="utf-8") as fh:
                data = yaml.safe_load(fh)
        except Exception as exc:
            logger.warning("Failed to parse %s: %s", yaml_path, exc)
            continue

        if not data or "atomic_tests" not in data:
            continue

        for idx, test in enumerate(data["atomic_tests"]):
            name = test.get("name", "").strip()
            description = test.get("description", "").strip()
            platforms = test.get("supported_platforms", [])
            executor = test.get("executor", {})
            executor_type = executor.get("name", "") if isinstance(executor, dict) else ""
            command = executor.get("command", "") if isinstance(executor, dict) else ""

            # Build an atomic_test_id in the format "T1059.001-0"
            atomic_test_id = f"{technique_id}-{idx}"

            source_url = (
                f"https://github.com/redcanaryco/atomic-red-team/blob/master"
                f"/atomics/{technique_id}/{technique_id}.yaml"
            )

            results.append({
                "technique_id": technique_id,
                "index": idx,
                "atomic_test_id": atomic_test_id,
                "name": name,
                "description": description,
                "platforms": ", ".join(platforms) if isinstance(platforms, list) else str(platforms),
                "executor_type": executor_type,
                "command": command[:4000] if command else None,  # cap at 4k chars
                "source_url": source_url,
            })

    logger.info("Parsed %d atomic tests total", len(results))
    return results


# ---------------------------------------------------------------------------
# Public API
# ---------------------------------------------------------------------------


def import_atomic_red_team(db: Session) -> dict:
    """Download and import Atomic Red Team tests as TestTemplates.

    Parameters
    ----------
    db : Session
        Active SQLAlchemy database session.

    Returns
    -------
    dict
        Summary with keys ``created``, ``skipped_existing``,
        ``yaml_files_parsed``, ``total_tests_parsed``.
    """
    tmp_dir = tempfile.mkdtemp(prefix="aegis_atomic_")
    try:
        zip_bytes = _download_zip()
        atomics_dir = _extract_zip(zip_bytes, tmp_dir)
        parsed_tests = _parse_yaml_files(atomics_dir)
    finally:
        # Always clean up
        shutil.rmtree(tmp_dir, ignore_errors=True)
        logger.info("Cleaned up temp directory %s", tmp_dir)

    # Pre-load existing atomic_test_ids for dedup
    existing_ids: set[str] = {
        row[0]
        for row in db.query(TestTemplate.atomic_test_id)
        .filter(TestTemplate.atomic_test_id.isnot(None))
        .all()
    }

    created = 0
    skipped = 0

    for item in parsed_tests:
        if item["atomic_test_id"] in existing_ids:
            skipped += 1
            continue

        template = TestTemplate(
            mitre_technique_id=item["technique_id"],
            name=item["name"][:500] if item["name"] else f"Atomic Test {item['atomic_test_id']}",
            description=item["description"][:2000] if item["description"] else None,
            source="atomic_red_team",
            source_url=item["source_url"],
            attack_procedure=item["command"],
            platform=item["platforms"],
            tool_suggested=item["executor_type"] if item["executor_type"] else None,
            atomic_test_id=item["atomic_test_id"],
            is_active=True,
        )
        db.add(template)
        existing_ids.add(item["atomic_test_id"])
        created += 1

    db.commit()

    # Count distinct YAML files by technique_id
    yaml_files_count = len({t["technique_id"] for t in parsed_tests})

    summary = {
        "created": created,
        "skipped_existing": skipped,
        "yaml_files_parsed": yaml_files_count,
        "total_tests_parsed": len(parsed_tests),
    }

    logger.info(
        "Atomic Red Team import complete — created=%d, skipped=%d, "
        "yaml_files=%d, total_tests=%d",
        created, skipped, yaml_files_count, len(parsed_tests),
    )

    # Audit log (system action)
    log_action(
        db,
        user_id=None,
        action="import_atomic_red_team",
        entity_type="test_template",
        entity_id=None,
        details=summary,
    )

    return summary