Processing Stix Taxii Feeds

Processing STIX/TAXII Feeds

When to Use

Use this skill when:

• Onboarding a new TAXII 2.1 collection from a government feed (CISA AIS, FS-ISAC) or commercial provider
• Validating that ingested STIX bundles conform to the OASIS STIX 2.1 specification before import
• Building automated pipelines that parse STIX relationship objects to reconstruct campaign context

Do not use this skill for proprietary vendor feed formats (Recorded Future JSON, CrowdStrike IOC lists) that require vendor-specific parsers rather than STIX processing.

Detection Gaps & Validation

• Confidence is optional and unscaled: STIX 2.1 confidence (0-100) is frequently absent, and producers map it differently (some use Admiralty/None-Low-Med-High bands, some omit it). Do not treat missing/low confidence as "ignore" or a high value as "block" - normalize each feed's scale on ingest and default missing values conservatively rather than to 100.
• Dedup and correlation gaps: the same indicator arrives from CISA AIS, an ISAC, and a commercial feed with different IDs and markings; without dedup by pattern value you inflate counts and may block on a single low-quality source. Key on the observable, not the STIX id.
• Silent data loss: missing pagination (next/as_pages) and clock skew on added_after drop objects at interval boundaries - use UTC with a 5-minute overlap and reconcile object counts per poll to detect gaps.
• Expired/revoked indicators: honor valid_until and revoked: true - pipelines that ingest the pattern but ignore these fields keep blocking on retired infrastructure. Filter both before routing to SIEM/firewall.
• How to confirm a bundle is usable: check spec_version (2.0 vs 2.1 schemas differ), confirm required indicator fields and pattern syntax parse, and resolve relationship refs (source_ref/target_ref) actually exist in the bundle before reconstructing campaign context.

Prerequisites

• Python 3.9+ with stix2 library (pip install stix2) and taxii2-client library
• Network access to TAXII 2.1 server endpoint with valid credentials
• Target TIP or SIEM with import API (MISP, OpenCTI, or Splunk ES)

Workflow

Step 1: Discover TAXII Server Collections

from taxii2client.v21 import Server, as_pages

server = Server("https://cti.example.com/taxii/",
                user="apiuser", password="apikey")
api_root = server.api_roots[0]
for collection in api_root.collections:
    print(collection.id, collection.title, collection.can_read)

Select collections relevant to your threat profile. CISA AIS provides collections segmented by sector (financial, energy, healthcare).

Step 2: Fetch STIX Bundles with Pagination

from taxii2client.v21 import Collection
from datetime import datetime, timedelta, timezone

collection = Collection(
    "https://cti.example.com/taxii/api1/collections/<id>/objects/",
    user="apiuser", password="apikey")

# Fetch only objects added in the last 24 hours
added_after = datetime.now(timezone.utc) - timedelta(hours=24)
for bundle_page in as_pages(collection.get_objects,
                             added_after=added_after, per_request=100):
    process_bundle(bundle_page)

Step 3: Parse and Validate STIX Objects

import stix2

def process_bundle(bundle_dict):
    bundle = stix2.parse(bundle_dict, allow_custom=True)
    for obj in bundle.objects:
        if obj.type == "indicator":
            validate_indicator(obj)
        elif obj.type == "threat-actor":
            upsert_threat_actor(obj)
        elif obj.type == "relationship":
            link_objects(obj)

def validate_indicator(indicator):
    required = ["id", "type", "spec_version", "created",
                "modified", "pattern", "pattern_type", "valid_from"]
    for field in required:
        if not hasattr(indicator, field):
            raise ValueError(f"Missing required field: {field}")
    # Check confidence range
    if hasattr(indicator, "confidence"):
        assert 0 <= indicator.confidence <= 100

Step 4: Route Objects to Consuming Platforms

Map STIX object types to destination systems:

• indicator objects → SIEM lookup tables and firewall blocklists
• malware objects → EDR threat intelligence library
• threat-actor / campaign objects → TIP for analyst context
• course-of-action objects → Security team wiki or SOAR playbook triggers

Use TLP marking definitions to enforce sharing restrictions:

for marking in obj.get("object_marking_refs", []):
    if "tlp-red" in marking:
        route_to_restricted_platform_only(obj)

Step 5: Publish Back to TAXII (Bi-directional Sharing)

# Add validated local intelligence back to shared collection
new_indicator = stix2.Indicator(
    name="Malicious C2 Domain",
    pattern="[domain-name:value = 'evil-c2.example.com']",
    pattern_type="stix",
    valid_from="2025-01-15T00:00:00Z",
    confidence=80,
    labels=["malicious-activity"],
    object_marking_refs=["marking-definition--34098fce-860f-479c-ae..."]  # TLP:GREEN
)
collection.add_objects(stix2.Bundle(new_indicator))

Key Concepts

Tools & Systems

• stix2 (Python): Official OASIS Python library for creating, parsing, and validating STIX 2.0/2.1 objects
• taxii2-client (Python): Client library for TAXII 2.0/2.1 server discovery, collection enumeration, and object retrieval
• MISP: Open-source TIP with native TAXII 2.1 server and client; MISP-TAXII-Server plugin for publishing MISP events
• OpenCTI: CTI platform with built-in TAXII 2.1 connector; supports STIX 2.1 import/export natively
• Cabby: Legacy Python TAXII 1.x client for older government feeds still on TAXII 1.1

Common Pitfalls

• Ignoring spec_version field: STIX 2.0 and 2.1 have incompatible schemas (2.1 adds confidence, object_marking_refs at bundle level). Always check spec_version before parsing.
• No pagination handling: TAXII servers cap responses at 100–1000 objects per request. Missing pagination (via next link header) causes silent data loss.
• Clock skew on added_after: Server and client time misalignment causes missed objects at interval boundaries. Use UTC exclusively and add 5-minute overlap windows.
• Storing raw STIX blobs without indexing: Storing bundles as opaque JSON prevents querying by indicator type or campaign. Parse into relational or graph database.
• Sharing TLP:RED content inadvertently: Automated pipelines must filter marking definitions before routing to any shared platform or SIEM with broad analyst access.