Emerging Cybersecurity Threats and Challenges
Recognise AI-written phishing and deepfake impersonation, apply call-back verification to a suspicious request, and choose layered defences for connected IoT devices.
When the boss's voice is not the boss
At 4:40 on a Friday afternoon, Sameer Joshi's phone rings with an unfamiliar number. The voice is unmistakably Vikram Deshmukh's, down to the way he says "Sameer, listen": he is stuck in a meeting in Mumbai and needs ₹8.5 lakh paid to a new supplier before the bank closes at 5. It is not Vikram. It is a deepfake, audio or video generated by artificial intelligence (AI), computer software trained on large amounts of data to produce realistic text, speech or images, to imitate a real person. Attackers have always adapted to new technology, and cheap, powerful AI tools are the latest change. Emerging threats are new or fast-growing kinds of attack that make yesterday's warning signs less reliable, and defenders need new habits to meet them.
AI-written phishing loses the old clues
Phishing is a fake message, usually an email or SMS, that tries to trick you into clicking a link, sharing a password or paying money. For years the easiest clue was poor writing: odd grammar, wrong spellings, a greeting like "Dear valued user". AI-assisted attacks remove that clue. An attacker can ask an AI tool to write a polite, fluent email in English, Hindi or Marathi, in the tone of Tulsi Health's real supplier, and produce a hundred different versions in a minute, each one slightly changed so that email filters see them as unrelated.
AI also helps attackers do their homework. Public posts, the company website and staff profiles on job sites reveal who manages accounts, which vendor supplies the billing software and when Vikram is travelling. A message that mentions Rathi Softech by name and refers to last week's actual software update feels genuine because the details are real. The defence is to stop judging a message by how well it is written and judge it by what it asks you to do.
Fluent writing proves nothing any more. Ask what the message wants you to do, and whether you can confirm it through a channel you already trust.
Deepfakes and impersonation
Impersonation means pretending to be someone the victim trusts. It is old, but deepfakes make it far more convincing. A few minutes of someone's recorded voice, from a video talk, a panel discussion or even voice notes, can be enough for today's tools to produce new sentences in that voice. Video deepfakes on a call are harder to make but improving quickly. Vikram has spoken at pharmacy association events, and those recordings are public.
The voice is perfect, but the request carries every classic warning sign. A good process catches what the ear cannot.
The lesson is uncomfortable: you cannot reliably tell a good deepfake by listening. So the defence must not depend on your ears. It depends on process. Call-back verification means ending the call and phoning the person back on the number you already have saved, never a number the caller gives you. A code word, a private phrase agreed in person and never written in email, lets a real Vikram prove himself in seconds. And a two-person rule, where any new payee or any transfer above, say, ₹1 lakh needs approval from a second person, means a single fooled employee cannot send the money alone. Like a bank locker that needs two keys, the rule protects the person holding one of them.
Sameer's Friday call is below as a transcript, along with a second call on Monday. Flag the parts that worry you, decide whether each call is a scam, and see how the call-back and code word settle the question either way.
IoT: the forgotten computers in every store
The Internet of Things, or IoT, means everyday objects that contain a small computer and connect to a network: smart TVs, doorbells, fitness bands. Every Tulsi Health store has two CCTV cameras and a smart temperature monitor in the medicine fridge, because insulin and vaccines must stay between 2 and 8 °C. The monitor sends readings to the pharmacist's phone. Each of these devices is a computer on the store network, yet nobody thinks of them that way.
Two small devices, a default password, old software and one shared network: three separate weaknesses an attacker only needs one of.
IoT security risks come from how these devices are made and forgotten. Many arrive with default passwords such as "admin", printed in manuals anyone can find online, and nobody changes them. Their software is rarely updated. They often sit on the same network as the billing laptops, so a hijacked camera becomes a doorway to more valuable machines. In 2016 a piece of malware called Mirai took over hundreds of thousands of cameras and home routers that still used default passwords and used them together to flood websites off the internet. For a pharmacy there is a quieter danger too: a tampered monitor that keeps showing 4.6 °C while the fridge warms, and spoiled insulin reaching a patient.
The defences are ordinary but must actually be done. Keep a list of every connected device. Change every default password. Update the device software or replace devices the maker no longer supports. Put cameras and monitors on their own separate network so they cannot reach billing laptops. Turn off remote access from the internet unless it is truly needed. Choose the defences for one store below within the budget, then run the attacks and see which layer stops each one.
Evolving attacks and the defender's dilemma
Attack methods keep changing because attackers adapt to whatever defences become common. When companies kept good backups, ransomware gangs started stealing data before encrypting it. When two-step login spread, some attackers began sending endless login approval requests until a tired employee pressed "Approve". Next year will bring tricks nobody has named yet, which is why habits such as checking through a second channel matter more than any list of today's scams.
Neha's team must guard every inbox, laptop and device every day. An attacker needs to find just one that was missed.
This imbalance is called the defender's dilemma: defenders must protect everything, all the time, while an attacker needs only one gap, once. It is like a fielding side in cricket that must cover every part of the ground, while the batter needs one gap to the boundary. No defender wins by being perfect, so the realistic answer is to make that single gap less damaging: layers of defence so one failure is not fatal, detection so an intruder is noticed quickly, tested backups so recovery is possible, and staff trained to pause and check. Defenders can also share what they learn, so an attack spotted at one company is recognised at the next, which slowly turns the dilemma around.