Built by the US Navy to Hide Spies: The Secret History of Tor and Onion Routing

The most-used anonymity network on the planet was not built by privacy activists. It was built by the United States military — specifically by mathematicians and computer scientists at the U.S. Naval Research Laboratory, who needed a way for intelligence officers to use the open internet without broadcasting that they worked for the government. The technology they invented is called onion routing, and the software that grew out of it is called Tor.

But here is the paradox that explains everything about how Tor works and why it exists in the form it does today: a secret communications tool used only by spies is worthless, because anyone watching the network can see that everyone on it is a spy. To hide its own users, the Navy's anonymity system had to be handed to journalists, activists, criminals, researchers, and ordinary people the world over. This article traces that history precisely — who built it, when, and why — and then does the thing most "Tor vs VPN" articles refuse to do: explain honestly what each one actually hides.

A spy network with one fatal flaw

Imagine an intelligence agency builds a private, encrypted network and lets only its own field officers use it. The encryption is flawless; no one can read the messages. There is still a catastrophic problem. Anyone who can observe internet traffic — a hostile government's telecom operator, for example — can see who connects to that network. They don't need to break the encryption. The mere fact that a computer in a foreign capital is talking to the "U.S. government anonymity system" is the entire secret, exposed.

This is the core insight at the heart of Tor's design. Anonymity is not a property of one person; it is a property of a crowd. You cannot hide in a group of one. The technical name for this is the anonymity set: the larger and more diverse the population of users, the harder it is to single out any individual within it. A military intelligence officer browsing the web is only invisible if a student in Brazil, a journalist in Turkey, and a privacy hobbyist in Germany are all using the exact same network at the exact same time, indistinguishable from one another.

A private network used only by spies is a neon sign that reads "spy here." The only way to hide the government's traffic was to bury it inside everyone else's.

Tor's own creators later wrote a paper making this point explicit, with the memorable title "Anonymity Loves Company." The usability of an anonymity system and its security are not separate concerns — they are the same concern. A tool that nobody else uses cannot protect anyone, including the people who built it. That is why the Navy could not keep onion routing classified and locked away. To protect its intended users, it had to give it to the world.

Born at the Naval Research Laboratory

Onion routing was conceived in the mid-1990s at the U.S. Naval Research Laboratory (NRL) in Washington, D.C. The three researchers credited with inventing it are mathematician Paul Syverson and computer scientists David Goldschlag and Michael Reed. Their goal was concrete and unglamorous: protect U.S. intelligence communications traveling over public networks, so that the origin and destination of sensitive traffic could not be reconstructed by an observer.

The first public description of the work appeared in 1996, in a paper titled "Hiding Routing Information," presented at an information-hiding workshop. The team also filed for a patent; U.S. Patent 6,266,704, "Onion routing network for securely moving data through communication networks," was assigned to the U.S. Navy and granted in 2001. The early prototype — sometimes called generation-zero onion routing — even ran its proof-of-concept on a single machine, more a research demonstration than a global network.

The name is a literal description of the technique. A message is wrapped in successive layers of encryption, like the layers of an onion, and each relay it passes through peels off exactly one layer — learning only enough to pass the message to the next hop, and nothing more. No single relay ever sees both who you are and what you are doing.

Onion routing in plain English

Here is what actually happens when you load a page over Tor. Your client software builds a path — called a circuit — through three volunteer-run relays scattered around the world. Your traffic is encrypted in three nested layers before it ever leaves your computer.

• The entry (or guard) relay sees your real IP address — it knows who you are — but because of the layered encryption, it has no idea what site you are visiting or what you are sending.

• The middle relay sees neither end. It knows only that it received data from the guard and must pass it to the exit. It is a blindfolded courier handing off a sealed package.

• The exit relay peels off the final layer and sends your request to the destination website. It sees what you are doing — the destination and any unencrypted content — but it sees the middle relay's address, not yours. It has no idea who you are.

The security of the system comes from this separation of knowledge. The piece that knows your identity (the guard) is blind to your activity. The piece that knows your activity (the exit) is blind to your identity. To unmask you, an adversary would generally need to control or observe both the entry and exit of your circuit and correlate the timing of the traffic — a real but difficult attack. By default Tor also rotates to new circuits over time, so you are not pinned to one path.

This is the structural difference that matters most when we compare Tor to a VPN later: trust in Tor is distributed across three independent parties, none of whom can deanonymize you alone. No single operator holds the whole picture.

From Navy project to nonprofit: the birth of the Tor Project

The version of the technology that most people use today — "Tor," originally an acronym for The Onion Router — is a second-generation redesign. Around 2002, NRL's Paul Syverson teamed up with two outside developers, Roger Dingledine and Nick Mathewson, to rebuild onion routing into something practical, deployable, and global. The Tor network's first relays went live in October 2002, and the foundational design paper, "Tor: The Second-Generation Onion Router," was published at the USENIX Security Symposium in 2004.

Crucially, the project went open source. The Navy released the Tor code under a free and open license, which meant anyone could read it, audit it, run a relay, and verify there was no hidden backdoor — an essential property for a tool whose entire value depends on users trusting it. In 2004 the Electronic Frontier Foundation (EFF) stepped in to fund continued development, helping move Tor out of a purely military lineage and into the civil-liberties world.

In December 2006, the developers incorporated The Tor Project, Inc. as a 501(c)(3) nonprofit research-and-education organization, with Dingledine, Mathewson, and Syverson among its founders. The Tor Project has since been funded by a mix of sources that includes U.S. government grants (the National Science Foundation, the State Department's internet-freedom programs, and what is now the U.S. Agency for Global Media), private foundations, and individual donations. You can read the project's own account of its mission and software at the Tor Project.

Tor vs. VPN: two different kinds of trust

Most "Tor vs VPN" comparisons frame the question as which one is "more private," as if privacy were a single dial. That framing is wrong. The honest difference is who you have to trust, and what they can see.

When you use a commercial VPN, your traffic is encrypted from your device to the VPN provider's server, which then forwards it to the wider internet using the provider's IP address. This hides your activity from your internet service provider and from the websites you visit (they see the VPN's IP, not yours). But notice what it does not do: the VPN provider itself sits in the middle and can see both your real IP address and every destination you connect to. You have not removed the single point of trust — you have moved it from your ISP to your VPN company. Everything rests on that one provider's honesty, its logging policy, and its resistance to legal pressure.

Tor was designed specifically to eliminate that single trusted party. Its three-relay structure means no one entity — not the guard, not the exit, not the website — knows both who you are and what you are doing. That is the trade Tor makes:

• VPN = single-provider trust. One company sees your identity and your destinations. Fast, simple, good for hiding traffic from your ISP and from sites — but only as trustworthy as that one company.

• Tor = distributed trust. Three independent relays, no central operator, no account, no payment trail. Much stronger against deanonymization — but slower, and with its own distinct weak point at the exit.

• Different jobs. A VPN is a privacy-and-access tool with a trusted operator. Tor is an anonymity system designed to remove the need to trust any operator at all.

The exit-node problem the listicles skip

Here is the detail that breezy comparison articles tend to omit. Because the Tor exit relay peels off the last layer of Tor's encryption to deliver your request to the destination, the exit operator can see whatever you send in the clear. If you log into a site over plain, unencrypted HTTP, the exit relay can read your username and password. Tor hides who you are from that relay, but it does not magically encrypt your content end-to-end with the website.

This is not theoretical. In 2007, security researcher Dan Egerstad set up a handful of Tor exit relays and harvested working email login credentials for dozens of embassies and government accounts around the world — not by breaking Tor, but simply by reading the unencrypted traffic that careless users were pushing through his exit nodes. The lesson stands today: anyone can volunteer to run an exit relay, including a hostile one, so transport encryption to the destination (HTTPS) still matters enormously when using Tor.

By contrast, a VPN's "exit" is the provider you chose and (ideally) vetted, not a random anonymous volunteer. Neither model is strictly safer — they fail in different ways. A VPN concentrates risk in a known party; Tor distributes it across unknown ones. Which is acceptable depends entirely on what you are defending against.

Who Tor actually protects — and the dark-web problem

Strip away the mythology and Tor's real-world user base is broad and mostly mundane. Investigative journalists use it to research stories and protect sources. Major news organizations and human-rights groups run Tor-based whistleblower submission systems — the open-source SecureDrop platform, used by outlets including newspapers and magazines, relies on Tor's hidden-service technology. Activists and ordinary citizens in countries with heavy censorship use it to reach the open web. Police and intelligence agencies use it for exactly the original NRL reason: to do their work without advertising their organization's IP addresses. Edward Snowden's leaked NSA materials in 2013 even included an agency slide deck candidly titled "Tor Stinks," complaining that the network was effective enough that the NSA could not reliably deanonymize all of its users on demand.

And then there is the reputation problem. The same hidden-service feature that lets a newspaper host an anonymous tip line also lets illegal marketplaces operate — most infamously the Silk Road drug market, shut down in 2013. This is the source of Tor's lurid "dark web" image. But the proportion matters: the overwhelming majority of Tor traffic is people anonymously browsing the ordinary web. Onion services (sites with .onion addresses reachable only through Tor) are a small fraction of activity, and they range from criminal markets to mirrors of mainstream news sites and major platforms that run them precisely to serve readers in censored regions.

Why governments fund the thing they also try to break

The strangest fact in Tor's history is that the U.S. government simultaneously bankrolls it and tries to defeat it — and both behaviors are rational. Different parts of a government have different missions. Internet-freedom and foreign-policy programs want a strong, popular anonymity network, because it helps dissidents and journalists under authoritarian regimes communicate, and because — back to the founding paradox — those same agencies' own operatives are only hidden if millions of other people are using the same tool. A weak Tor with few users protects no one, including them.

Meanwhile, signals-intelligence and law-enforcement arms of the same government have a mandate to surveil specific targets, and for them a strong anonymity network is an obstacle. So they research attacks against it. There is no contradiction once you accept that "the government" is not one actor with one goal. Tor's continued existence, partly on government money, is the institutional expression of the very paradox it was born from: anonymity for the few requires anonymity for the many.

Takeaway: choose by threat model, not by hype

The question "Is Tor or a VPN more private?" has no answer because it is the wrong question. The right question is: what are you hiding, and from whom? Decide based on your threat model.

1. Hiding browsing from your ISP, or unblocking content, with good speed and one trusted operator? A VPN fits that job — provided you actually trust the provider, because it can see everything you do.

2. Defeating identification when no single party should ever link your identity to your activity — for a journalist, source, dissident, or researcher? Tor's distributed-trust design is what it was built for, accepting slower speeds and the need for HTTPS to guard against hostile exit relays.

3. Facing a powerful adversary that can watch large parts of the network? Understand that no consumer tool is a magic cloak; operational discipline (what you log into, what you reveal) often matters more than the tool itself.

The history of Tor is ultimately a lesson in counterintuitive design. The most powerful privacy network we have exists because a military intelligence lab understood that secrecy and crowds are not opposites — that the only way to hide a few important people was to protect everyone. Onion routing did not leak out of the Navy by accident. It was released on purpose, because that was the only way it could ever work.

Frequently Asked Questions

Who created Tor?

Onion routing, the technology Tor is built on, was invented in the mid-1990s by Paul Syverson, David Goldschlag, and Michael Reed at the U.S. Naval Research Laboratory. The modern Tor software was then developed starting around 2002 by Syverson together with Roger Dingledine and Nick Mathewson, who later co-founded the nonprofit Tor Project in December 2006.

Did the US government really build Tor, and the onion routing at the Naval Research Lab?

Yes. Onion routing was a U.S. Naval Research Laboratory project, funded by the Navy and later DARPA, originally to protect government and intelligence communications online. The code was deliberately released as open source, and the Tor Project still receives part of its funding from U.S. government internet-freedom programs alongside foundations and donors.

Why was Tor made public if it was built for spies?

Because anonymity requires a crowd. A network used only by government agents would instantly identify everyone on it as a government agent. To hide its intended users, the tool had to be opened to journalists, activists, and ordinary people, so that no single user stands out within a large, diverse population.

What is the difference between Tor and a VPN in terms of history and design?

Tor grew from a military research project into a nonprofit anonymity network that spreads trust across three independent relays, so no single party knows both who you are and what you do. A VPN is a commercial service where one provider routes your traffic and can see both your identity and your destinations. Tor removes the single trusted party; a VPN concentrates trust in one company you have to vet.

Is Tor the same as the dark web?

No. Tor is an anonymity network for using the regular internet privately, and the vast majority of its traffic is ordinary web browsing. The "dark web" refers to .onion hidden services reachable only through Tor, which are a small subset of activity and range from criminal markets to legitimate whistleblower systems and censored-region news mirrors.

Is Tor safe to use, and what is the exit-node risk?

Tor strongly protects your identity, but the final relay — the exit node — can read any traffic you send unencrypted, because it strips Tor's last layer to reach the destination. Anyone can run an exit relay, including a malicious one; in 2007 a researcher captured embassy credentials this way. Using HTTPS sites mitigates this, since your content stays encrypted to the destination even at the exit.

Should I choose Tor or a VPN?

Decide by threat model, not by which is labeled "more private." A VPN suits hiding traffic from your ISP or unblocking content with one trusted operator and good speed. Tor suits situations where no single party should ever link your identity to your activity, such as journalism, sources, or evading censorship, at the cost of speed.