Imagine a security analyst at an invented Hyderabad insurance firm, scrolling through her queue on a quiet Thursday. Four hundred alerts, almost all noise: a login from a new city, a rare process, a spike in traffic. Then one alert reads differently. Someone, from an ordinary employee laptop, has tried to connect to a file server called FIN-ARCHIVE-OLD. She checks the asset list. That server does not exist as a real system. It was created eight months ago specifically so nobody would ever need to connect to it.

She does not need to debate whether it is a false positive. Nobody has a business reason to touch that server. She isolates the laptop within minutes.

The principle: legitimate users have no reason to be there

Most detection works by finding bad behaviour inside a sea of normal behaviour, which produces ambiguity. Deception flips this. A honeypot is a system or service deliberately set up to be attractive to attackers and useless to everyone else. Because no genuine activity should ever reach it, any interaction is suspicious by definition. The ratio of signal to noise is very high.

Kinds of decoys

Honeypots vary mainly by how much they imitate.

  • Low-interaction: software that pretends to offer a service, such as a login prompt or an open port, and records connection attempts. Cheap and safe to run, but a determined human will quickly see it is fake.
  • High-interaction: a real, fully functioning system that is deliberately monitored and isolated. It lets you observe what an attacker actually does, but it takes effort and carries risk if isolation is poor.
  • Internet-facing research honeypots: exposed to the world to watch the background noise of scanning and commodity attacks. They are useful for learning and sharing intelligence, less so for protecting a specific company.
  • Internal decoys: fake servers, fake shares and fake accounts planted inside your own network, which is where the Hyderabad example sits. These detect an intruder who is already in and looking around.

Canary tokens: decoys without a server

A canary token is a tiny tripwire that sends an alert when someone uses it. Examples: a document named “Salary_Revision_Master” that contacts a tracking address when opened; a fake cloud access key that is never used legitimately, so any attempt to use it fires an alert; a web link in a hidden part of an internal wiki; a database record with a made-up customer that should never be queried; a fake user account in the directory whose name looks like a privileged administrator.

The beauty is cost. A fake key planted in a code repository takes minutes. If the repository is later copied by a thief or scraped, and someone tries the key, you learn that the repository leaked, roughly when, and from which address. Free and open tools exist for generating such tokens, and some commercial platforms manage them at scale.

Where decoys disappoint

Honesty about limits matters here.

A decoy that looks fake gets ignored. Real servers have patterns: hostnames matching a naming scheme, believable services, traffic, users, files with realistic dates. Many deployments fail because the honeypot is the only machine in the subnet with no history. Also, authorised scanning by your own security team or vulnerability scanner will trigger alerts on a honeypot, which creates noise unless you exclude those sources. Be careful as well with a high-interaction system: if it can reach production networks, you have built a stepping stone for the attacker. Isolate it fully.

There are legal and policy questions too. Observing an attacker on your own decoy is generally fine, but deliberately luring people, or using decoys to collect data from third parties, can be sensitive. Check with legal counsel before deploying anything internet-facing, and note that in India the CERT-In directions on incident reporting may apply when a real compromise is discovered, honeypot or not.

Finally, a decoy is only useful if its alerts are routed to someone who will act. An alert from FIN-ARCHIVE-OLD that sits unread in a mailbox does nothing. Write the response in advance: who gets paged, what is isolated first, how you confirm.

Starting small

You do not need a deception platform on day one. Plant three or four tokens in places an intruder would look: a fake admin credential in a script folder, a decoy file share named as something tempting, a fake record in a sensitive database. Document each, alert to the SOC, and test that each actually fires. Quiet systems are good news, but you should know they are quiet because nothing is happening, not because the wiring is broken.

Names and scenario in the example are illustrative.

Leave a Reply

Trending