Picture a Tuesday afternoon at a bank’s support desk. A caller says she is Meera Nair and that she has lost her phone. She is calm, polite, and slightly rushed. She wants her registered mobile number changed so she can receive OTPs on her new SIM. The agent, Karthik, asks for her name, date of birth and the last four digits of her account number. She answers instantly. Karthik, relieved by a pleasant call after three angry ones, updates the number.

Two days later the real Meera finds her account emptied. The caller had her name, birth date and last four digits from a leaked database and a phishing message. Karthik followed the script and still did the wrong thing. We will come back to what he should have done, because the answer says a lot about how authentication is designed.

Authentication is not identification

Identification means “who does this caller say they are”. Authentication means “can they prove it”. A name and account number only identify. Proof needs something an impostor would struggle to have. Security people group proof into three families: something you know (a PIN, a password, an answer), something you have (a phone that receives an OTP, a card, a hardware token) and something you are (a voice print, a fingerprint). Strong processes combine two families, which is why you see an OTP on top of a security question.

The weakness in Karthik’s check was that all three answers were things a stranger could find out. Name, date of birth and partial account number appear in forms, KYC records, old couriers’ labels, data dumps. Knowledge that is widely shared is not knowledge that proves anything. Good practice is to ask for details that are not printed on documents and not easily searched: a recent transaction amount, a value only the customer would see in the app, or a PIN set specifically for phone banking.

Risk decides how hard to ask

If every question needs full verification, genuine customers get annoyed and call times rise. If nothing does, fraud walks in. Mature processes tier the request by risk. Reading out the store’s opening hours needs no verification. Giving a balance needs basic checks. Changing a mobile number, email address, address or credentials, adding a beneficiary, or issuing a duplicate card are the actions that fraudsters want, because they let them take over the account. These need step-up authentication: extra questions, an OTP sent to the number already on file, a call back, or a rule that the change must be done through a channel that is already authenticated, such as the logged-in app.

Back to Karthik. The right move was to notice that the request itself was the red flag: change of contact detail plus “I lost my phone” is the classic account takeover story. The process should have blocked a phone-based change altogether, or forced verification in person, by video, or through a documented route with a cooling-off period. The point is that the safeguard belongs in the process, not in an agent’s instinct on a good day.

How fraudsters work the agent

Social engineering is the term for manipulating people rather than breaking systems. The common tools are not clever, they are human:

  • Urgency. A flight in an hour, a hospital bill, a child in trouble.
  • Sympathy. Crying, apologising, a crying baby in the background.
  • Authority. “I am calling from head office”, “I am a senior manager”, “I will report you”.
  • Partial knowledge. Quote some true details to sound legitimate, then ask the agent to fill in the rest.
  • Probing. Call several times, with different agents, until one is lenient. Each failure teaches the caller which question was the hard one.

The agent’s best defence is a rule that does not bend. Saying “I would love to help, and I can only do this after verification” is not rude. It is the same sentence the customer would want said to a stranger asking about their account.

OTPs and a warning about them

OTPs are widely used, and they cut both ways. A one-time code sent to the registered number proves possession of that number. Companies often tell customers never to read an OTP out to anyone, and rightly so, since fraudsters ring up pretending to be the bank. If your own process asks customers to read an OTP to the agent, you are teaching the habit that scammers exploit. Designs differ: some firms have the customer type the code into an IVR, some verify by app notification, some use voice biometrics instead. If you are designing or auditing a process, check that what agents ask for matches what the public is warned about. Contradictory messages confuse honest customers.

SIM swap deserves a mention too. If a fraudster gets a duplicate SIM, they receive the OTP. That is why high-risk actions sometimes pair the OTP with another signal, such as a recent SIM change flag from the operator, where available, or a cooling-off period.

Failing a customer gracefully

Genuine customers fail verification too. They forget which email they registered, or the account is in a spouse’s name. The agent should not hint at answers, should not say which field was wrong, and should not let the caller guess repeatedly. A small limit on attempts, then a route to an alternative: branch visit, app login, or a call-back to the registered number. Document the failed attempt. A pattern of failed calls on the same account is something a fraud team can use.

Third parties are the hardest case. A son calling for an elderly father who is ill is plausible, and the process must still be clear about what is allowed: usually no account information unless the account holder has authorised that person in a recorded way. Firm and kind can go together.

In India, this also sits within data-protection duties. The Digital Personal Data Protection Act, 2023 puts obligations on organisations to safeguard personal data, and sector regulators such as the RBI have their own expectations for banks and payment firms. A wrongly disclosed account is not only an operational error; it can become a reportable incident. How exactly the obligations apply depends on your client, your role as processor or fiduciary, and the rules notified so far, so your compliance team is the authority there.

Figures and names in the examples are illustrative.

Leave a Reply

Trending