Threat feed

CRAFTY FALCON payment API fraud campaign

CRAFTY FALCON invokes victims' own payment APIs to drain Indian financial firms and, in its third wave, alters business records to redirect legitimate payments

Posted · Spectrum Delta Research Team

Summary

CRAFTY FALCON, an AI-assisted financial threat actor designated by India's CERT-In CSIRT-Fin (A/2), steals funds from Indian financial firms by entering through exposed non-production environments and invoking the victims' own payment and disbursement APIs to move money directly. CERT-In has documented the campaign across three advisories between 9 October and 20 November 2026, naming the ARTEX, PentAGI, Cairn and CyberStrikeAI offensive AI frameworks in the 4 November escalation. The 20 November advisory adds a third wave beyond finance: intruders alter business records and approval relationships inside service-provider payment workflows so unauthorized transactions pass normal approval checks, and remediated entities are deliberately re-targeted, timed to weekends and holidays. Documented tradecraft also blends command execution into AWS Systems Manager and Lambda channels, bypasses IP allow-lists through shared third-party fetcher services, and validates flows with micro-value test transactions before large transfers. No public indicators exist; fingerprints and detection rules are restricted to TLP:AMBER advisories, and the actor's origin and sponsorship remain unknown.

Breakdown

Countries

  • India

Targets

  • Industry: Non-banking financial companies
  • Industry: Digital lending and fintech platforms
  • Industry: Payment aggregators and payment service providers
  • Industry: Cryptocurrency exchanges
  • Industry: Technology service providers supporting payment workflows

Technical exposure

  • Internet-exposed non-production environments sharing credentials with production
  • Payment and disbursement APIs callable with stolen aggregator credentials
  • IP allow-lists trusting shared third-party fetcher services
  • Cloud management services reachable with compromised credentials
  • PHP and Node.js applications with file-upload or deserialization flaws
  • Business applications where payment records can be modified undetected

Adversary

CRAFTY FALCON · AI-assisted financial threat actor designated by CERT-In CSIRT-Fin (A/2); origin and sponsorship unknown

Procedures (6)

  1. Web shells planted in internet-exposed non-production apps

    • T1190Exploit Public-Facing Application
    • T1505.003Server Software Component: Web Shell

    Logs

    • Web application firewall logs · Exploit attempt then shell: file-upload or deserialization payloads against internet-facing UAT, staging or testing hosts, followed by PHP or Node.js web shell traffic from those hosts
    • File integrity monitoring · File created: web shell files (new or renamed .php or .js) in application deployment directories of non-production systems; collection requires FIM on web roots and staging paths
  2. Cloud management services run attacker commands and deploy files

    • T1059Command and Scripting Interpreter

    Logs

    • CloudTrail · Management events: SendCommand or StartSession (Systems Manager), UpdateFunctionCode (Lambda), and creation of roles, users or access keys; interactive remote-management sessions and serverless code updates outside approved automation principals
  3. Non-production credentials reach production systems

    Logs

    • Firewall or proxy traffic logs · Connection: non-production hosts or subnets (UAT, staging, test naming) reaching production services, databases or management ports
    • Identity provider audit logs · Authentication or token use: shared service accounts first seen against non-production, then production resources; API tokens used from unusual source IPs, user agents or geographies
  4. Payment APIs called with stolen credentials from fetcher IPs

    Logs

    • Payment platform audit logs · API calls to disbursement, fund-transfer or merchant-settlement endpoints using aggregator, gateway or partner-bank credentials; micro-value test transactions (about one rupee) to new beneficiaries followed by high-value transfers
    • Web application firewall logs · Payment-endpoint API requests originating from shared third-party fetcher IP pools (content-acceleration or app-store review fetchers) rather than known client infrastructure
  5. Business records altered so payments pass approval checks

    • T1565.001Data Manipulation: Stored Data Manipulation

    Logs

    • Business application audit logs · Update events: payment instructions, beneficiary details or approval-relationship records changed by service-provider accounts, followed by transactions that pass normal maker-checker approval; values diverging from independently held authoritative records
  6. Rapid variants of functionally identical attack scripts created

    Logs

    • File integrity monitoring · Bursts of newly created commented scripts and rapid successive variants of functionally identical scripts in web roots, temporary directories and staging paths; a pattern CERT-In links to AI-generated attack tooling

Sources

Questions, detections or sightings of CRAFTY FALCON payment API fraud campaign? Bring them to the community.