How Attackers Bypassed NBFC Payout Controls — and What CERT-In Recommends
How attackers abused trusted NBFC payout APIs, what CERT-In's October and November 2026 advisories reveal, and six controls lenders should test now.
- Published
- Reading time
- 8 min
- Filed under
- API Security · CERT-In · NBFC · Payment Security
On this page (20)
The recent attacks on Indian lenders expose a fundamental weakness in payment security: a trusted API connection is not the same as an authorised transaction.
An attacker does not necessarily need to compromise a bank's systems to steal money through its fund-transfer API. Compromising the lender's infrastructure, obtaining valid API credentials and bypassing the controls that should authorise a payment can be enough.
That is the attack pattern described in two critical advisories issued by India's Computer Emergency Response Team (CERT-In) in October and November 2026.
For NBFCs, digital lenders and payment companies, the immediate question is no longer simply whether their APIs have vulnerabilities. It is whether their systems can distinguish a legitimate disbursement from a technically valid but unauthorised instruction.
What CERT-In reported
On 9 October 2026, CERT-In published Advisory CIAD-2026-0046, describing attacks against NBFCs, digital lending platforms, fintech companies and payment providers.
The attackers compromised internet-facing applications, exposed APIs or privileged systems. After obtaining payment-related credentials, they invoked fund-transfer APIs directly from infrastructure that banks or payment partners already recognised as trusted.
The fraudulent transactions bypassed checks ordinarily performed by lending applications, including loan creation, KYC, sanction approval, beneficiary verification and transaction limits.
The advisory described dummy transaction references, rapid high-value transfers through IMPS and NEFT, and funds reaching mule accounts across multiple banks.
On 4 November, CERT-In issued a second critical advisory, CIAD-2026-0058, identifying the campaign as CRAFTY FALCON.
The update reported continued attacks, including against organisations that had already undertaken remediation. CERT-In also described attacks timed around weekends and holidays, when monitoring and response coverage may be reduced.
The November advisory added detail about AI-assisted offensive tooling, compromised cloud-management services, exposed testing environments, modified client applications and techniques used to bypass IP-based trust controls.
Together, the advisories describe an active threat to the systems that initiate, approve and transmit financial transactions.
How a valid API call becomes a fraudulent transfer
Consider a lending platform that disburses approved loans through a partner bank.
Under normal conditions, the loan management system creates a loan record, verifies the borrower, approves the sanction, validates the beneficiary and instructs the bank to release the funds.
The bank's API receives the final instruction from an authorised integration server.
The security problem arises when the bank-facing API accepts an instruction based primarily on the identity of that server or its credentials, without independently verifying that the underlying transaction has passed the necessary business checks.
CERT-In's reported attack patterns illustrate how this can be exploited.
1. Initial access
Attackers exploit an exposed application, a deprecated API, an inadequately protected administrative interface or credentials left in source code and configuration files.
2. Access to payment infrastructure
They move through the environment, potentially using shared credentials or connectivity between testing and production systems, and obtain the credentials needed to invoke payment APIs.
3. Direct API invocation
Instead of creating a loan through the intended application workflow, the attacker calls the fund-transfer API directly, potentially from the lender's own authorised infrastructure.
4. Unauthorised disbursement
The instruction may contain a fabricated loan reference, an unapproved beneficiary or an amount that should exceed the permitted transaction limits.
5. Rapid fund movement
In the attacks described by CERT-In, funds were transferred to attacker-controlled accounts, sometimes after small test transactions established that the transfer path worked.
6. Evidence destruction
The attackers also attempted to delete logs, interfere with endpoint protection and conceal their activity.
The important distinction is that an authenticated API call is not necessarily an authorised business transaction.
A bank may correctly authenticate an integration server while the lender's controls fail to establish whether that particular payment should exist.
What changed in November: faster attacks and more paths to production
CERT-In's November update makes the problem more difficult for conventional security operations.
The advisory describes attackers using AI-assisted tools to accelerate reconnaissance, exploitation and the generation of tailored scripts. It also identifies techniques that exploit weaknesses in cloud administration, application deployment and the separation of test and production environments.
Three developments are especially significant.
Testing environments became entry points. Internet-accessible UAT and staging systems provided routes into production where credentials, tokens or network access were shared.
IP allow-lists were not sufficient. CERT-In described abuse of shared third-party fetcher infrastructure, demonstrating why recognising a source IP address cannot replace strong application identity and transaction authorisation.
The attackers returned. Some previously affected organisations were targeted again after incomplete remediation.
These observations reinforce a basic principle: blocking an individual exploit or rotating a single credential does not necessarily remove the attack path.
The entire route from initial access to fund movement must be examined.
Six actions lenders should prioritise now
CERT-In's recommendations extend across applications, infrastructure, credentials, payment controls, monitoring and incident response. The following six priorities bring the most relevant measures together for organisations operating payout and disbursement APIs.
1. Establish which APIs and environments can reach production
Start with an inventory of every API and system involved in disbursements, refunds, merchant settlements and other fund movements.
Identify unused endpoints, deprecated integrations and internet-facing UAT or staging environments.
Check whether non-production systems share production credentials, databases, API keys or direct network paths.
An API that is no longer used by the application but remains accessible may still provide an attacker with an entry point.
Immediate test: Can a user or process operating in UAT reach production payment systems or obtain credentials that work in production?
2. Verify the caller, not merely its IP address
A trusted server address does not establish that the software issuing a request is authorised to move money.
Review long-lived static tokens, shared credentials and broad API permissions. Where appropriate, use short-lived credentials, mutually authenticated connections, cryptographically signed requests and controls against replay.
The receiving service should verify the identity and authority of the calling application for the requested operation.
Immediate test: If an attacker steals an API key or gains access to an authorised integration server, what transactions can they initiate without further approval?
3. Enforce business approvals at the payment stage
This is the most important control for the attack pattern described by CERT-In.
A payout should be rejected unless it can be linked to a valid, approved transaction.
The payment stage should verify the sanction record, beneficiary, amount, approval state and applicable limits. Those checks should not depend solely on the user interface or an earlier application step that an attacker can bypass.
CERT-In recommends cryptographically binding the stages of a transaction so that a debit-stage API cannot be invoked independently of the required validation.
Maker-checker controls, beneficiary verification and transaction ceilings should also operate at the relevant payment boundary.
Immediate test: Can the payout API accept a fabricated loan reference, a changed beneficiary, a modified amount or a previously approved instruction submitted again?
4. Protect credentials, source code and deployment paths
Review how bank API credentials and other payment secrets are stored, accessed, rotated and revoked.
Search source repositories, configuration files and CI/CD pipelines for embedded secrets. Restrict production deployment privileges and review changes to sensitive code and configurations.
The November advisory also highlights the abuse of cloud-management functions and modified application packages.
An organisation must therefore consider both how an attacker obtains a payment credential and how that credential can subsequently be used.
Immediate test: Can a compromised application, build process or cloud-management identity read payment credentials or execute unauthorised code in the disbursement environment?
5. Detect suspicious transfers and reconcile them quickly
Prevention matters, but lenders also need to detect transactions that should never have occurred.
CERT-In recommends monitoring fund movements for unusual transaction volumes, high-value transfers, new beneficiaries and patterns involving small test transactions followed by larger payments.
Monitoring should be capable of triggering a hold or review where suspicious activity is detected.
Reconciliation between internal transaction records and bank or payment-provider records should happen sufficiently quickly to identify unauthorised transfers while intervention remains possible.
The October advisory also recommends centralised logging with retention of at least 180 days.
Immediate test: Would monitoring flag a ₹1 transfer to a new beneficiary followed by multiple high-value transfers outside normal business hours? How quickly would the reconciliation process identify transactions without corresponding approved records?
6. Prepare to contain a compromised payout integration
An incident-response plan should identify who can suspend a payment API, disable compromised accounts, revoke credentials and coordinate with banking partners.
It should also cover preservation of logs and forensic evidence, investigation of persistence mechanisms and verification that the attacker has been removed before normal operations resume.
Because CERT-In observed attacks during weekends and holidays, the response process must work when normal staffing and approval chains are unavailable.
Affected organisations should follow applicable incident-reporting obligations and the reporting guidance in the CERT-In advisories, including coordination with CERT-In/CSIRT-Fin and relevant financial-fraud reporting channels.
Immediate test: If unauthorised payouts begin at 2 a.m. on a public holiday, who can stop them, how do they do it, and how long does that decision take?
Why an ordinary API penetration test may not be enough
Traditional API security testing examines vulnerabilities such as broken object-level authorisation, excessive permissions, authentication weaknesses, injection and business-logic abuse.
These remain essential.
But the payout-API attacks raise additional questions that may span several applications, infrastructure components and operational teams.
An API may pass authentication testing while still accepting a payment that has no valid sanction.
A transaction-limit control may work in the lending application but not at the final payment stage.
A monitoring platform may collect every API request while failing to recognise that a transfer has no corresponding loan record.
That is why a complete assessment must examine both the APIs and the controls connecting them.
The relevant evidence can include API requests and responses, approval records, transaction logs, credentials and permissions, reconciliation results, alert rules and incident-response procedures.
The objective is to determine whether the control chain works from beginning to end, and not merely whether each individual component appears secure.
What should an NBFC do this month?
A practical starting point is to identify every bank integration capable of moving funds and answer five questions for each one:
- Can a payout be initiated without a valid sanction or approved source transaction?
- Can the beneficiary or amount be altered after approval?
- Can a stolen credential or compromised integration server initiate unauthorised transfers?
- Would monitoring and reconciliation detect a suspicious transfer promptly?
- Can the organisation suspend the integration and investigate an incident outside normal business hours?
The answers should be supported by configuration evidence, transaction records and controlled testing, not assumptions about how the application is intended to work.
CERT-In recommends regular vulnerability assessment and penetration testing, including API security testing. Its public advisories provide security recommendations; they should not be confused with a universal requirement that every NBFC obtain an identical audit report.
Where a bank or regulator has issued specific assessment instructions, those requirements should be established separately.
Assess the entire payout journey
Security Brigade provides technical API security testing and a broader Payout API Security Audit covering the controls surrounding fund transfers.
The assessment examines how applications, integration servers, credentials, transaction approvals, beneficiary checks, monitoring and reconciliation work together. Technical testing is used to validate relevant attack paths, with findings documented alongside the supporting evidence and remediation status.
For lenders, the key question is straightforward:
If an attacker can call your payout API from a server you already trust, what prevents that call from becoming a transfer you never approved?
Explore Security Brigade's Payout API Security Audit or learn about our API Security Testing service.
Primary sources
- CERT-In Advisory CIAD-2026-0046: “Threat Actors Exploiting Payment and Disbursement APIs for Unauthorized Fund Transfers in BFSI Entities”, 9 October 2026
- CERT-In Advisory CIAD-2026-0058: “Escalation of AI-Driven Attacks on BFSI Entities by Threat Actor CRAFTY FALCON”, 4 November 2026
About the authors
Chintan J
CISO & Director — Security Advisory
Oversees Security Brigade's cybersecurity advisory practice, helping regulated enterprises meet RBI, SEBI, CERT-In, and IRDAI compliance mandates. Previously held senior security leadership roles across BFSI.
Security Brigade Editorial Team
Regulatory and Technical Research · Security Brigade
The editorial team reads the primary sources so you do not have to. Every regulatory piece here is written against the circular, Direction or standard itself and cites the paragraph, and every technical piece is reviewed by the practice lead who runs that kind of engagement.
Continue reading
All articles →RBI Cybersecurity Directions for NBFCs: Which Chapter Applies to Your Layer
RBI/DoS/2026-27/461 does not apply uniformly. Chapter III binds Base Layer NBFCs under ₹500 crore and Core Investment Companies, Chapter IV binds Base Layer at ₹500 crore and above, and Chapter V binds Middle Layer and above. The bands are mutually exclusive.
RBI Cybersecurity Framework in 2026: What Replaced It, and What Banks Must Do Now
The framework you are looking for was repealed on 31 July 2026, along with 627 other circulars. Six new Directions replaced it, one per class of regulated entity, all in force on issuance. What changed, what carried forward, and the cadence most guidance is stating incorrectly.
DPDP Rule 6: the seven security safeguards, and the runway to May 2027
Rule 6's seven minimum safeguards commence 13 or 14 May 2027. What each limb must produce, and the two that need a budget.