API Security

₹3.3 Crore in 87 Minutes: App and API Security Lessons for BFSI

5 min read Updated

A week ago, a CISO at a large financial institution told me about a tough conversation in his boardroom following the October 2026 NBFC cyberattacks (where one Hyderabad lender reportedly lost ₹3.3 crore across 183 transfers in just 87 minutes). His board had asked him a painfully simple question: “We invested heavily in risk-based vulnerability management to prioritize and close critical gaps. How can attackers exploit known weaknesses to move crores in minutes, while our teams take weeks to discover and fix those same exposures?”

From the board’s perspective, the reaction was: “What? How did our risk dashboard not catch this?” After all, banks and large non-banking financial companies (NBFCs) run mature security and vulnerability management systems that scan assets, score severity, and track remediation SLAs.

When he asked how I would answer them, I was blunt: “Risk-based vulnerability management alone is not going to cut it anymore. The board has a point. We need to apply virtual patches as soon as we discover a vulnerability, and give the DevOps team time to fix the code. You and I both know it still takes weeks to test and apply patches in production. And that is only half the problem: our controls assume that anything arriving with the right keys from a trusted source is legitimate.”

That speed gap and that misplaced trust are precisely what played out across Indian NBFCs in October, and escalated on November 4 when CERT-In warned of the AI-assisted CRAFTY FALCON campaign (Advisory CIAD-2026-0058). And it is about to become every financial institution’s problem; those lenders just experienced it first.

(For the attack timeline and API payloads used, read Vivek Gopalan’s companion technical breakdown [here].)

The Problem in One Sentence

Most enterprise vulnerability programs are built around known CVEs that take weeks to patch, while modern API heists chain unpatched or misconfigured servers with business-logic flaws in custom code that never get a CVE in the first place.

Look at how the two stages CERT-In observed actually worked, because the detail matters:

Stage 1

Initial access through exposed, unpatched apps and APIs

Attackers did not need a zero-day. According to CERT-In’s October 9 advisory (CIAD-2026-0046), they broke in through unpatched application server middleware, misconfigured internet-facing servers, unrestricted file-upload flaws, and deprecated or unauthenticated APIs left online. Notice the word “unpatched”: a fix already existed, but the server was either waiting in a patching queue or living on a forgotten “shadow” or “zombie” endpoint that nobody was scanning.

Stage 2

Fraud through non-CVE business-logic gaps

Once inside, attackers harvested partner API keys that were hard-coded in source code or stored in plaintext in configuration files and called loan-disbursal APIs directly from the lender’s own whitelisted server. In digital lending, the API is the transaction. The disbursal API paid out crores on dummy reference IDs because it trusted the server and never checked whether an approved loan actually existed.

This is also why traditional risk-based vulnerability management dashboards missed the risk. Somewhere along the way, our industry let a lookup key (the CVE ID) become the proxy for risk. Most risk-scoring tools prioritize known CVEs and infrastructure weaknesses. Because a broken loan-disbursal workflow, a missing authorization check, or a shadow API never receives a CVE, routine scanning rarely flags it. Mature programs do find these flaws through penetration testing and dynamic application testing, but periodically and in a separate report, not on the risk dashboard. Too often, no ticket is created and no SLA clock ever starts.

Your risk dashboard stays green the whole time. And in banking cybersecurity, green-but-wrong is much worse than red.

Why the Old Remediation Pipeline Breaks at AI Speed

Now go back to Stage 1. Why were those doors still open when the attackers showed up? Because in most enterprises, the remediation chain looks like this:

Scan > Prioritize > Ticket > SLA > Wait weeks for a developer sprint.

Before anyone writes it off, this was a sound design for years. Waiting weeks for a code fix is a property of the software pipeline, not a failure of the people running it.

Then offensive AI arrived. As CERT-In’s November 4 advisory highlights, CRAFTY FALCON uses publicly available AI tools to probe non-production environments, generate exploit scripts on the fly, and re-attack firms even after initial remediation. When attackers can move ₹3.3 crore out of a lender in 87 minutes, waiting three weeks for a developer sprint is like bringing a calendar to a stopwatch fight.

For the flaws your scanners do catch, discovery is no longer the bottleneck; what matters is how fast you block exploitation. CERT-In’s November 4 advisory itself recommends WAF virtual patching as an interim control.

From Detection to Autonomous Protection

When I talk about autonomous vulnerability mitigation at Indusface, I mean something very practical: closing exploitable gaps in hours without waiting for code releases. AppTrana and SwyftComply AI run as one continuous loop across four connected steps:

Step 1

Discover (External and East-West): You cannot protect an API you forgot existed. AppTrana maps documented, shadow, and zombie APIs in live traffic, while our new East-West API discovery (launching mid-November 2026) uses lightweight server sensors to map internal APIs and outbound calls to partner banks.

Step 2

Test (AI Scale with Human Judgment): Scanners look for signatures; attackers look for loopholes. AI-assisted penetration testing, backed by years of our own pentesting intelligence, probes multi-step loan workflows for business-logic flaws, with every finding validated by human experts.

Step 3

Protect (Autonomous Virtual Patching): Here is where we break the sprint bottleneck. SwyftComply AI turns validated flaws into expert-verified custom security policies at the edge, enforcing virtual patches, schema checks, and velocity limits in hours. The result is a clean vulnerability report you can hand to auditors within 72 hours.

Step 4

Monitor (24/7 Expert Oversight): Because attackers love long weekends, our 24/7 team watches live traffic and pushes proactive rules as new threat campaigns emerge.

“Show Me the Evidence”: Regulators Move from Guidance to Scrutiny

The RBI is pressing banks to prove cyber controls work in practice and to audit API links with NBFC and fintech partners (If you connect to a bank, that audit is coming to your doorstep next: an RBI advisory dated November 3 reportedly requires banks to have NBFC API clients audited by CERT-In-empaneled auditors, with existing clients given until February 3, 2027). Meanwhile, CERT-In advises a full API inventory, retirement of unused endpoints, and centralized logs retained for at least 180 days, while the DPDP Act will turn API data leaks into reportable personal data breaches from May 2027.

What to do this week

5 Questions for BFSI Leaders

  1. Do we have a live map of our external and East-West APIs, including legacy apps, test servers, and outbound partner-bank links?
  2. Are lending and payout flows tested for business-logic abuse, verifying every payout against a sanctioned loan?
  3. Can we block a critical flaw at the edge in hours, rather than waiting weeks for a code release?
  4. Will velocity controls automatically halt abnormal payout bursts in real time?
  5. Are edge protections paired with core banking safeguards like hardware security modules for keys, maker-checker rules, and real-time reconciliation?

So, Where Does This Leave You?

Your security program cannot wait weeks for a software release to act. Block the exposure at the edge in hours, and let developers fix the code on their normal schedule. If your gap between discovery and protection is already zero, congrats, and please tell me how you did it. If it is not, you have just found your blind spot.

Could Your APIs Be Exposed to Similar Attacks?

Request a BFSI API Security Assessment with Indusface to map your external and East-West API exposure, uncover business-logic risks, and protect critical applications in hours with AppTrana and SwyftComply AI.

Sources

Indusface
Indusface

Indusface is a leader in autonomous application security, protecting thousands of business-critical web, API, and AI applications for global enterprises across 45+ countries. As AI-driven exploitation outpaces traditional patching cycles, its flagship AppTrana platform continuously discovers vulnerabilities and autonomously mitigates exploitable risks through AI-powered virtual patching. With SLA-backed, zero-false-positive protection and 24/7 expert oversight, AppTrana defends organizations against API abuse, DDoS attacks, malicious bots, AI threats, and zero-day exploits, keeping them protected long before code fixes ship.