Denial-of-Service and Network Disruption
Distinguish a denial-of-service attack from a genuine traffic rush and choose defences such as CDNs, filtering and rate limiting that keep a service available.
The festival evening the app stopped answering
On the Saturday before Diwali, Tulsi Health runs a festival sale in its app. At 7:40 pm Rohan Iyer's phone starts buzzing: the app takes forty seconds to open, then shows an error. On a normal sale evening the app receives about 1,500 requests a second and can handle about 2,000. Tonight 30,000 a second are arriving, from thousands of different addresses, and almost none of them are buying anything. Nothing has been stolen and no record has been changed, yet Mrs Lata Gokhale cannot reorder her blood pressure tablets. This is a denial-of-service (DoS) attack: an attempt to make a service unavailable to the people who should be able to use it, usually by overloading it.
Two words will help throughout. A server is a computer that waits for messages and answers them, like the machine that runs the Tulsi Health app and its database. A request is one such message, for example "show me my past orders" or "add this strip to my cart". Every time Mrs Gokhale taps a button in the app, her phone sends a request across the internet and waits for the server's answer.
A crowd blocking the shop entrance
Network availability means that a network and the services on it are reachable and respond in reasonable time whenever the people who rely on them need them. It is the "A" in the confidentiality, integrity and availability triad, and it is the property a DoS attack targets. Every system has a capacity, a limit on how much work it can do at once: so many requests a second for a server, so much data a second for an internet connection.
Picture the Kothrud store with one pharmacist at the counter. Suppose two hundred people walk in together, each asks one question, buys nothing, leaves and comes straight back. The store is not broken and nothing is stolen, but Mrs Gokhale cannot get through the crowd to the counter. Traffic overload is the network version: more requests arrive than the system can handle, so answers slow down, queues fill, and new requests are dropped. The attacker does not need to break in at all, only to send more than the system can take.
A denial-of-service attack does not break the shop. It fills the entrance so that real customers cannot get in.
Attackers can aim at different limits, just as a crowd could block the road, the door or the pharmacist. Some floods simply send more data than the internet connection can carry. Some open conversations with the server and never finish them, so the server's list of open conversations fills up. Others ask for the most expensive work the app can do, such as a search across every medicine, again and again.
Each flood targets a different resource. The defence that helps depends on which one is being exhausted.
One attacker or ten thousand: DoS and DDoS
A DoS attack from a single computer is fairly easy to stop: once you spot the one address sending the flood, you block it. A distributed denial-of-service (DDoS) attack sends the flood from thousands of devices at once. Those devices usually belong to a botnet, a large group of hijacked devices that an attacker controls remotely. Cheap internet cameras, home routers still using their factory passwords and infected computers all end up in botnets, and their owners rarely know.
Blocking one sender is easy. Blocking ten thousand ordinary-looking senders without blocking real customers is the hard part.
DDoS is harder because each device sends only a little, so no single address looks dangerous, and the addresses belong to ordinary homes, some of them in the very cities Tulsi Health serves. Some attacks add spoofing: they send small questions to public servers on the internet with the victim's address forged as the sender, so the servers send large answers to the victim. This is called reflection, and when the answer is much bigger than the question, amplification. The motives vary: criminals demanding money to stop ("pay ₹5 lakh or the app goes down again tonight"), protesters making a point, or attackers creating noise to distract the security team from a break-in elsewhere.
Busy evening or attack?
Not every overload is an attack. A festival sale, a viral post or the first day of the month when many customers reorder can all produce a genuine rush, sometimes called a flash crowd. The difference shows in the details. Real customers arrive soon after an announcement, browse different pages, fill carts and pay. Attack traffic tends to repeat the same request, comes at odd hours or from unexpected places, and almost never ends in an order. Telling them apart matters, because the right response to a rush is to add capacity, while the right response to an attack is to filter it.
Sort the situations below the way Neha would on a busy night. Look for signs of real customer behaviour, such as orders and varied pages, before deciding.
Absorb, filter, limit and prepare
No single defence stops every flood, so organisations stack several, each catching what the one before it missed. Rate limiting caps how many requests one address or one account may send each second, say 20. A real customer sends one or two, so they never notice, while a single noisy sender is cut down at once. Against a large botnet it helps much less, because each of thousands of devices can stay politely under the limit.
A content delivery network (CDN) is a service with servers in many cities that keep copies of a website's pages and answer visitors from the nearest one. Its combined capacity is far larger than any single company's server, so it can absorb a flood that would bury Tulsi Health's own machine. Traffic filtering, often bought from the same provider or from the internet service provider, inspects incoming traffic and drops what matches the attack, such as known botnet addresses, malformed messages or one request repeated endlessly, before it reaches the app. It works like the society gate guard who turns away the crowd while letting known residents in.
Each layer removes part of the flood, so the app itself sees mostly real customers.
Preparation matters as much as technology. Neha keeps a short plan: what normal traffic looks like, who calls the provider, how to switch on stronger filtering, and what message customers see while the problem is fixed. Extra capacity, such as servers that are added automatically when load rises, helps with rushes and small attacks, but it costs money, and an attacker with a large botnet can usually send more than a company can afford to buy.
Try the defences yourself on the festival evening below. Start with the mission: the attack and the sale are fixed, so you have to find the combination of defences that keeps real customers served, and notice which defence does almost nothing against thousands of devices.