Aegis/backend/app/services/intel_service.py

"""Automated threat-intelligence scan service.

Searches public security RSS feeds for mentions of MITRE ATT&CK technique
IDs and names.  New findings are stored as :class:`IntelItem` records and
the related technique is flagged for review.

This is an **MVP** implementation — it queries a small set of well-known
RSS feeds and parses them with the standard-library :mod:`xml.etree`
parser.  No LLMs or paid APIs are used.
"""

import logging
import re
import defusedxml.ElementTree as ET
from datetime import datetime

import requests as _requests
from sqlalchemy.orm import Session

from app.models.intel import IntelItem
from app.models.technique import Technique
from app.services.audit_service import log_action

logger = logging.getLogger(__name__)

# ---------------------------------------------------------------------------
# Public security RSS feeds
# ---------------------------------------------------------------------------

RSS_FEEDS: list[dict[str, str]] = [
    {
        "name": "CISA Alerts",
        "url": "https://www.cisa.gov/cybersecurity-advisories/all.xml",
    },
    {
        "name": "NIST NVD CVE",
        "url": "https://nvd.nist.gov/feeds/xml/cve/misc/nvd-rss.xml",
    },
    {
        "name": "SANS ISC",
        "url": "https://isc.sans.edu/rssfeed.xml",
    },
    {
        "name": "BleepingComputer",
        "url": "https://www.bleepingcomputer.com/feed/",
    },
    {
        "name": "The Hacker News",
        "url": "https://feeds.feedburner.com/TheHackersNews",
    },
    {
        "name": "Krebs on Security",
        "url": "https://krebsonsecurity.com/feed/",
    },
]

# Timeout for each feed request (seconds)
_FEED_TIMEOUT = 15

# Minimum technique name length for name-based matching
# Short names ("Kill", "BITS") produce too many false positives
_MIN_NAME_LEN = 8


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


def _fetch_feed(url: str) -> list[dict[str, str]]:
    """Download and parse an RSS/Atom feed, returning a list of entries.

    Each entry is a dict with keys ``title``, ``link``, and ``description``.
    Returns an empty list on any error so the scan can continue.
    """
    try:
        resp = _requests.get(url, timeout=_FEED_TIMEOUT, headers={
            "User-Agent": "AegisPlatform/1.0 IntelScan",
        })
        resp.raise_for_status()
    except Exception as exc:
        logger.warning("Failed to fetch feed %s: %s", url, exc)
        return []

    try:
        root = ET.fromstring(resp.content)
    except ET.ParseError as exc:
        logger.warning("Failed to parse feed %s: %s", url, exc)
        return []

    entries: list[dict[str, str]] = []

    # RSS 2.0 format: <channel><item>...
    for item in root.iter("item"):
        title_el = item.find("title")
        link_el = item.find("link")
        desc_el = item.find("description")
        entries.append({
            "title": title_el.text.strip() if title_el is not None and title_el.text else "",
            "link": link_el.text.strip() if link_el is not None and link_el.text else "",
            "description": desc_el.text.strip() if desc_el is not None and desc_el.text else "",
        })

    # Atom format: <feed><entry>...
    ns = {"atom": "http://www.w3.org/2005/Atom"}
    for entry in root.iter("{http://www.w3.org/2005/Atom}entry"):
        title_el = entry.find("atom:title", ns)
        link_el = entry.find("atom:link", ns)
        summary_el = entry.find("atom:summary", ns)
        link_href = ""
        if link_el is not None:
            link_href = link_el.get("href", "")
        entries.append({
            "title": title_el.text.strip() if title_el is not None and title_el.text else "",
            "link": link_href.strip(),
            "description": summary_el.text.strip() if summary_el is not None and summary_el.text else "",
        })

    return entries


def _build_patterns(technique: Technique) -> tuple[list[re.Pattern], list[re.Pattern]]:
    """Build regex patterns for a technique.

    Returns two lists:
    - ``id_patterns``:   MITRE ID patterns (high confidence, word-boundary matched)
    - ``name_patterns``: technique name patterns (lower confidence, long names only)
    """
    id_patterns: list[re.Pattern] = []
    name_patterns: list[re.Pattern] = []

    # MITRE ID with word boundaries so T1059 doesn't partially match T1059.001
    mitre_id_escaped = re.escape(technique.mitre_id)
    id_patterns.append(re.compile(rf"\b{mitre_id_escaped}\b", re.IGNORECASE))

    # Technique name — only for distinctly long names to reduce false positives
    if technique.name and len(technique.name) >= _MIN_NAME_LEN:
        name_escaped = re.escape(technique.name)
        name_patterns.append(re.compile(rf"\b{name_escaped}\b", re.IGNORECASE))

    return id_patterns, name_patterns


def _entry_matches(
    entry: dict[str, str],
    id_patterns: list[re.Pattern],
    name_patterns: list[re.Pattern],
) -> bool:
    """Return True if any pattern matches the entry's title or description."""
    text = f"{entry.get('title', '')} {entry.get('description', '')}"
    return any(p.search(text) for p in id_patterns + name_patterns)


def _entry_matches(entry: dict[str, str], patterns: list[re.Pattern]) -> bool:
    """Return True if any pattern matches the entry's title or description."""
    text = f"{entry.get('title', '')} {entry.get('description', '')}"
    return any(p.search(text) for p in patterns)


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


def scan_intel(db: Session) -> dict:
    """Run the intel scan across RSS feeds for known techniques.

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

    Returns
    -------
    dict
        Summary with keys ``new_items``, ``duplicates_skipped``,
        ``techniques_flagged``, ``feeds_checked``.
    """
    logger.info("Intel scan starting...")

    # 1. Load all active techniques
    techniques = (
        db.query(Technique)
        .order_by(Technique.mitre_id)
        .all()
    )
    logger.info("Scanning %d techniques against %d feeds", len(techniques), len(RSS_FEEDS))

    # 2. Pre-load all existing intel URLs for dedup
    existing_urls: set[str] = {
        row[0] for row in db.query(IntelItem.url).all()
    }

    # 3. Fetch all feeds once
    all_entries: list[tuple[str, dict[str, str]]] = []  # (feed_name, entry)
    feeds_ok = 0
    for feed in RSS_FEEDS:
        entries = _fetch_feed(feed["url"])
        if entries:
            feeds_ok += 1
        for entry in entries:
            all_entries.append((feed["name"], entry))

    logger.info("Fetched %d entries from %d/%d feeds", len(all_entries), feeds_ok, len(RSS_FEEDS))

    # 4. Match entries to techniques
    new_items = 0
    duplicates_skipped = 0
    techniques_flagged: set[str] = set()

    for technique in techniques:
        id_patterns, name_patterns = _build_patterns(technique)

        for feed_name, entry in all_entries:
            if not _entry_matches(entry, id_patterns, name_patterns):
                continue

            # Skip entries with no title (low-quality)
            if not entry.get("title", "").strip():
                continue

            url = entry.get("link", "").strip()
            if not url:
                continue

            # Dedup
            if url in existing_urls:
                duplicates_skipped += 1
                continue

            # Create IntelItem
            intel_item = IntelItem(
                technique_id=technique.id,
                url=url,
                title=entry.get("title", "")[:500],
                source=feed_name,
                detected_at=datetime.utcnow(),
                reviewed=False,
            )
            db.add(intel_item)
            existing_urls.add(url)
            new_items += 1

            # Flag technique for review
            if not technique.review_required:
                technique.review_required = True
                techniques_flagged.add(technique.mitre_id)

    # 5. Single commit
    db.commit()

    summary = {
        "new_items": new_items,
        "duplicates_skipped": duplicates_skipped,
        "techniques_flagged": len(techniques_flagged),
        "feeds_checked": feeds_ok,
    }

    logger.info(
        "Intel scan complete — new=%d, duplicates_skipped=%d, "
        "techniques_flagged=%d, feeds_checked=%d",
        new_items, duplicates_skipped, len(techniques_flagged), feeds_ok,
    )

    # 6. Audit log
    log_action(
        db,
        user_id=None,
        action="intel_scan",
        entity_type="intel_item",
        entity_id=None,
        details=summary,
    )
    db.commit()

    return summary