How to read Disclaimer

Part C Β· The deep and dark web

Chapter 14 Β· Surface, deep and dark: mapping the hidden internet

Technical intensity: Low Free preview

Learning outcomes

  1. You can define the surface web, the deep web and the dark web correctly, and you can sort any everyday online thing into the right layer β€” including the traps that fool most people.
  2. You can explain why famous size claims like "the dark web is 500 times bigger than the internet" are wrong, where the number actually comes from, and what honest measurements of Tor look like today.
  3. You can tell the history of anonymous networks in order: David Chaum's 1981 idea, the cypherpunk remailers, onion routing at the US Naval Research Laboratory, Tor's launch, Silk Road, and the takedown era.
  4. You can separate myth from measured reality about crime on the dark web, using published research rather than headlines.
  5. You can explain why defenders and investigators need this map at all, and (technical tier) install Tor Browser safely and visit legitimate onion services only.

Concepts

Three layers, one internet

There is only one internet. The "three layers" are not three separate networks. They are three levels of visibility β€” who can find a piece of content, and how.

The surface web is everything a search engine can reach, read and list. The news homepage you found this morning, Wikipedia articles, a company's public "About us" page. Search engines send out programs that follow links from page to page and record what they find. If a page can be found that way, it is surface.

The deep web is everything the search engines cannot or do not list. Not because it is criminal β€” because of four ordinary doors. Door one: a login. Your bank statement, your email inbox, your university's student portal. Door two: a query. The page does not exist until you ask for it β€” a library catalogue result, a court-case status page, a flight price. Door three: a paywall β€” journalism and academic journals you must buy. Door four: an explicit request not to be indexed, or simply having no links pointing to the page. An intranet, an internal wiki, a cloud drive shared with five colleagues.

The dark web is a small slice of the deep web that needs special software to reach at all, because both the visitor and the visited site hide their network location. In practice that software is mainly Tor; there are also smaller networks such as I2P and Hyphanet, covered in Chapter 19. Sites on Tor are called onion services and their addresses end in .onion. A browser without Tor simply cannot open them.

Now the sentence worth memorising: the dark web is a small part of the deep web, and the deep web is most of the web. "Deep" does not mean "dark". Your bank statement is deep web. It is also the most ordinary object in the world. Journalists, researchers and police organisations run legitimate onion services on the dark web; Chapter 16 collects them. Crime exists there too, and Chapters 21 to 26 examine it with numbers, not adjectives.

visible to search engines waterline = indexed? SURFACE β€” indexed pages DEEP β€” logins, databases, records, paywalls, intranets legal, normal, and larger than the surface by nature DARK β€” Tor onion services thin slice: roughly 600–700 thousand addresses seen per day (see below) Figure 1: an honest iceberg. Even this misleads β€” database rows cannot be counted as "pages".

Notice what the classic iceberg gets wrong. It draws the underwater mass as one dark, scary block. In truth the underwater mass is almost entirely ordinary β€” databases, account pages, records β€” and the dark web is a thin slice inside it, smaller by every honest measure. If a diagram cannot show that, it is marketing, not a map.

πŸ”§ Technical

In plain words: a search engine can only list pages its crawler programs can reach and read without logging in. Everything behind a login, a query box, a paywall or a "do not index" note stays out of the list. Invisible to Google is not the same as secret.

How indexing actually works, in four steps. A crawler (Googlebot is one) starts from known pages and follows every link it finds. It hands the text of each page to an index β€” a giant catalogue. When you search, the engine ranks matching index entries. Pages drop out of this pipeline when:

  • they sit behind a login the crawler has no account for;
  • they are generated only in answer to a query (the catalogue page for one book never exists as a stored page);
  • the site's robots.txt file or a page's noindex tag asks crawlers to stay out β€” sites do this deliberately;
  • no link anywhere points to the page, so the crawler never discovers it.

A site asking not to be indexed, in its own words:

# robots.txt β€” a public note websites leave for crawlers
User-agent: *           # applies to every crawler
Disallow: /private/     # please do not crawl anything under /private/

And the shape of a page that only exists after you ask for it β€” your own bank statement is built like this:

https://portal.example-bank.example/login            (door: login)
https://portal.example-bank.example/statement?month=2026-09&format=pdf
   (door: query β€” this page was assembled for you, on demand)

Honest measurement: how big is it, really?

You will meet the myth family in many costumes: the dark web is "500 times bigger" than the surface web, or holds "96 per cent of the internet", or is an ocean of hidden crime. The family resemblance is that nobody measured any of it. The famous multiplier traces to Bergman's 2001 study of the deep web β€” an estimate that public information in searchable databases was 400 to 550 times larger than the linked web of March 2000, in an era before modern search engines handled dynamic content at scale.[3] Three things make the recycled number meaningless today. It counted database contents, not web pages β€” comparing rows of a catalogue to pages of a newspaper. It described the deep web, not the dark web. And it is a quarter of a century old.

What can be measured is Tor itself, because Tor publishes its own statistics. In October 2026, Tor Metrics estimated roughly three million mean daily users of the network, with India around one hundred and thirty thousand users per day β€” sixth worldwide β€” and roughly 590,000 to 700,000 unique onion addresses seen per day.[5] These are extrapolations, and the technical block below shows you how to read and challenge them. For contrast: the surface web runs to billions of indexed pages, and the deep web cannot be counted at all in any comparable unit. By every honest yardstick, the dark web is the smallest of the three layers.

Even the onion-service count deserves a warning label. A 2025 peer-reviewed measurement study from IMDEA Networks and Universidad Carlos III de Madrid crawled hidden services for months and found an ecosystem that is highly unstable: many addresses are replicas of the same sites, many vanish shortly after appearing, and domains rotate constantly. Illicit services existed β€” the researchers reported previously unknown child-abuse sites to the Spanish police β€” but so did news mirrors, secure-communication platforms and privacy tools.[6] "Hundreds of thousands of addresses" does not mean hundreds of thousands of living, distinct, criminal websites. It means churn.

πŸ”§ Technical

In plain words: Tor publishes its own network statistics, and anyone can check them. The user numbers are careful estimates worked out from network traffic, not sign-ups or headcounts, so read them as ranges and trends, not exact truths.

Reading metrics.torproject.org like an analyst:[5]

  • Users. Tor cannot count people, so it estimates users from the directory requests clients make. The by-country table for late 2026 showed the United States at about 564,000 mean daily users, equal to 17.77 per cent β€” which implies a worldwide total of about 3.2 million. That total is our arithmetic from their table, not a figure Tor prints on the page; treat it as an order of magnitude. India stood near 134,000 (4.22 per cent, sixth).
  • Onion services. The chart of unique version-3 onion addresses seen per day showed roughly 590,000–700,000 in October 2026. The page states these are extrapolated from statistics that single relays report, used only when at least 1 per cent of relays report them.
  • Events and anomalies. The metrics pages annotate real-world events (for example, internet-shutdown periods visible in country data) and occasionally show implausible spikes that the Tor Project itself flags. Estimates are useful; scepticism is professional.
  • Reproducibility. Every chart offers its raw data for download, so an investigator can re-plot and cite exact values instead of screenshotting an infographic.

This is the discipline this book will reuse everywhere in Part C: a claim about the hidden internet is worth exactly as much as its measurement method.

A short, accurate history

The hidden internet did not appear with crime, and it was not built for crime. It grew out of a forty-year-old mathematics idea. The timeline below is the spine; later chapters put flesh on each bone.

  1. 1981 β€” the idea. David Chaum publishes "Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms" in the journal Communications of the ACM. He proposes the mix network: messages relayed through servers that shuffle and re-encrypt them so no single server sees both sender and receiver.[7] Every anonymity system in this book is a descendant of this paper.
  2. Early 1990s β€” the believers. The cypherpunks, a loose mailing-list movement of cryptographers and hobbyists, turn Chaum's mathematics into working tools: anonymous remailers that strip sender identity from email. The best-known service, anon.penet.fi in Finland, ran from 1993 until legal pressure closed it in 1996.[8]
  3. Mid-1990s β€” the navy's problem. Researchers at the US Naval Research Laboratory β€” Paul Syverson with Michael Reed and David Goldschlag β€” invent onion routing: wrapping traffic in layers of encryption and passing it through several relays, each of which peels one layer. The goal was protecting government communications, not hiding markets.[9]
  4. 2000 β€” Freenet. Ian Clarke releases Freenet (March 2000), a decentralised store designed to resist censorship: files live across volunteers' computers with no central server to seize. The original project is renamed Hyphanet in 2023 when its founder reused the name for a redesign.[10]
  5. 2002–2004 β€” Tor. Roger Dingledine and Nick Mathewson, working with Syverson, build the second-generation onion router: Tor, for "The Onion Router". The code is released openly and the network begins running with a handful of volunteer relays β€” about a dozen by the end of 2003. Their 2004 paper states the design choice that made Tor matter: the network is open to everyone, because anonymity only works when normal users blend in.[9][11] In 2003, separately, the I2P project also begins, growing out of the Freenet lineage.[12]
  6. 2006 β€” an institution. The Tor Project, Inc. becomes a non-profit research organisation, independent of the US Navy that funded the original work.[9]
  7. 2011 β€” Silk Road opens (February), combining Tor onion services with Bitcoin payments. It is the first large hidden market, and it becomes the template β€” and later, the cautionary tale β€” for everything in Chapters 21 and 25.
  8. 2013 β€” the seizure. The FBI arrests Ulbricht and seizes Silk Road (October).[1] The mass media discovers "dark web" as a word. Measurement research matures in parallel: Christin's audit of Silk Road in 2013, Soska and Christin's ecosystem study in 2015.[4]
  9. 2014–2019 β€” legitimacy and takedowns. Facebook runs an official onion service from 2014; the New York Times from 2017; the BBC mirrors its news on Tor in October 2019 for readers under censorship.[13] Meanwhile police dismantle market after market β€” AlphaBay and Hansa in 2017, Hydra in 2022 β€” stories Chapter 26 tells properly.
  10. 2020s β€” the shift. Much criminal activity moves off Tor to ordinary messaging platforms such as Telegram, chasing convenience. The dark web becomes quieter, smaller, and still watched. Chapter 23 covers the shift and what it means for investigators.
πŸ”§ Technical

In plain words: Tor sends your traffic through three volunteer computers, wrapped in three layers of encryption. Each computer peels one layer, so no single computer knows both who you are and where you are going.

The shape of one Tor circuit, drawn as text:

You ──[3 layers]──> Guard relay ──[2 layers]──> Middle relay ──[1 layer]──> Exit relay ──> Website

Guard   : peels layer 1, learns you exist, never learns your destination
Middle  : peels layer 2, knows neither end
Exit    : peels layer 3, talks to the website, never learns who you are
Website : sees the exit relay's address, not yours
  • The guard is chosen once and kept, to limit how many relays ever see your address.
  • When the destination is itself an onion service, the meeting happens entirely inside Tor: there is no exit relay and no exposure to the ordinary web at all. That is why the circuit display in Tor Browser shows three relays and no exit for onion sites.
  • This is the one-minute version. Chapter 18 opens every one of these boxes β€” how relays are chosen, how onion services introduce themselves, and where the design is genuinely weak.

Myths versus measured reality

Myth 1: the dark web is basically crime. Measurement keeps refusing to cooperate. The 2025 IMDEA survey found news mirrors, secure drop systems for sources, and privacy tools running alongside the markets β€” and found the illicit portion volatile and churn-heavy rather than a stable empire.[6] Christin's Silk Road audit β€” the most-cited measurement of a criminal market β€” totalled about 1.2 million dollars of monthly sales: serious, prosecutable, and comparable to a single mid-sized shop, not a parallel economy.[4] Crime on the dark web is real, and this book treats it seriously for eleven chapters. It is a slice, not the whole.

Myth 2: it is enormous. Roughly three million daily Tor users is fewer people than use one large Indian city's metro on a busy day; unique onion addresses are in the hundreds of thousands, mostly replicas and dead ends.[5][6]

Myth 3: anonymity makes people uncatchable. The takedown record says otherwise. Market after market has fallen β€” not usually by breaking Tor's mathematics, but through human operational mistakes: reused passwords, sloppy server configuration, leaked emails. Chapters 25 and 26 are the case studies.

Myth 4: it is useless or dangerous to know about. For a defender, the hidden internet is a working environment: leaks surface there, stolen credentials are traded there, and threat intelligence is collected there. Avoiding the topic leaves you blind to the part of the threat landscape where evidence actually appears.

Writers who have looked closely at this folklore β€” Jamie Bartlett in The Dark Net and Robert Gehl in Weaving the Dark Web β€” land on the same observation from different directions: the moral-panic framing sells headlines, and the measured picture is smaller, stranger and more human: ordinary people seeking privacy, some criminals, and a lot of empty churn.[8][14] Chapter 17 returns to the sociology; Chapter 16 to the legitimate uses.

Why defenders and investigators care

Five practical reasons, all developed later in Part C:

  1. Leaks surface below the surface. When data is stolen, it reappears on forums, paste sites and markets that search engines never list. Monitoring that space is a defensive discipline (Chapter 27).
  2. Credential dumps are deep-web objects. A dump of email-password pairs is a database, traded and searched, not a page Google lists. Breach-notification services exist to watch it for you (Chapter 15).
  3. Your own organisation may appear there. Company names, employee credentials, leaked documents. Finding that early is incident response; finding it late is a news article.
  4. Investigations end in evidence. Understanding how the networks work β€” and fail β€” is what turns a hunch into a case (Chapters 25, 26, 30).
  5. Fraud and finance intersect here. Carding, payment fraud and crypto cash-out all route through hidden services; a fintech-focused defender meets this economy sooner or later (Chapters 22, 24, 29).

This chapter is the map. The territory begins in Chapter 15 β€” with the deep web proper, where most investigative value actually lives.

Worked demo: one evening online, three layers deep

Take one ordinary person β€” call her Meera β€” and follow thirty minutes of her evening. At each step, ask the investigator's question: which layer is this in, and who can see it?

Meera searches for her bank's homepage. The search results page and the bank's homepage are surface web β€” indexed, public. She logs in. Her account page and the statement PDF she downloads are deep web: created for her after a query, behind a login, deliberately invisible to search engines. Nothing sinister has happened; the layer changed because the visibility rules changed.

She reads a news article about her bank β€” surface again. The article links to an analyst report that needs a subscription β€” deep, behind a paywall. Her employer's HR portal where she checks her payslip β€” deep, an intranet. Finally, the same newspaper's note for readers in censored countries, explaining its Tor mirror on an address ending in .onion β€” dark web, legitimate, and visited for exactly the same journalism.[13]

Now the defender's half. For each step, someone else could also observe: the search engine logs the query; the bank's servers log the login and the statement download; the paywall provider knows which article was fetched. Layers describe search-engine visibility, not secrecy from everyone. Keeping that distinction sharp is what stops investigators confusing "I cannot Google it" with "nobody can see it".

Layer audit of Meera's evening
1. Search results page ................ surface (indexed)
2. Bank homepage ....................... surface (indexed)
3. Account page + statement PDF ......... deep   (login + query)
4. News article ........................ surface (indexed)
5. Analyst report ...................... deep   (paywall)
6. HR portal / payslip ................. deep   (intranet, login)
7. Newspaper's .onion mirror ........... dark   (Tor only, legitimate)
Who else can see each step: engine logs (1), bank logs (2-3),
publisher logs (4-5), employer logs (6), Tor relays see encrypted
traffic only; the newspaper's server sees content served (7).
Surface web
Web content that search engines can reach, read and list in their indexes.
Deep web
Web content search engines do not index: behind logins, generated by queries, paywalled, or excluded by request. Legal, normal, and most of the web.
Dark web
The small part of the deep web that needs special software (mainly Tor) to reach, because visitor and site both hide their network location.
Onion service
A website reached through Tor at an address ending in .onion, designed so the site's location and the visitor's location stay hidden.
Onion routing
Passing traffic through several relays, each peeling one layer of encryption, so no single relay sees both origin and destination. Invented at the US Naval Research Laboratory in the mid-1990s.
Mix network
David Chaum's 1981 design: servers that batch, shuffle and re-encrypt messages to break the link between sender and receiver. The ancestor of all anonymity networks.
Tor
"The Onion Router" β€” the second-generation onion-routing network (public since the early 2000s) and today the main doorway to the dark web.
Tor Metrics
The Tor Project's public statistics service: users, relays and onion-service counts, published with their estimation methods.
Crawler and index
The programs and catalogues a search engine uses to discover and list pages. What the crawler cannot reach, the index cannot list.
Iceberg diagram
The recurring infographic of surface/deep/dark layers. Usually wrong in its proportions; useful only when drawn honestly.
Cypherpunks
The 1990s movement of hobbyists and researchers who built practical anonymity tools, such as anonymous remailers, from academic cryptography.

Summary

There is one internet and three levels of visibility. The surface web is what search engines list; the deep web is everything they cannot β€” logins, databases, paywalls, intranets β€” and it is most of the web; the dark web is a thin slice of the deep web that needs Tor, where both ends of a connection hide. The famous size myths trace to Bergman's 2001 deep-web study and never measured the dark web at all. Honest numbers come from Tor Metrics: in October 2026, roughly three million daily users and a few hundred thousand onion addresses β€” most of them churn and replicas, per the 2025 IMDEA measurement study. The history runs from Chaum's 1981 mix networks through cypherpunk remailers, naval onion routing, Tor's open launch, Freenet and I2P, to Silk Road and the takedown era β€” a history of privacy engineering, only lately shadowed by crime. Myths of a criminal empire dissolve against measurement; what remains is a small, watchable, partly criminal ecosystem that defenders cannot afford to ignore, because leaks, credentials and evidence surface there. The map is now yours; Chapter 15 walks into the largest layer, the deep web proper.

Lab β€” Core tier: classify the layer (no tools needed)

⚠ Authorized practice only β€” this exercise, like every exercise in this book, runs only on systems you own or are explicitly authorised to test. Read Legal & ethical use first.

Objective: sharpen the one skill this chapter teaches β€” deciding, with reasons, which layer something belongs to. Work on paper. Ten minutes. The items are deliberately mixed; several are traps.

Steps

  1. For each item below, write down your classification β€” surface, deep or dark β€” and one sentence of reason. Name the specific door where one applies (login, query, paywall, noindex/no-links, or special-software).
  2. For items 11 and 12, add a second line: who else can see this, apart from the user?

Items

  1. The Wikipedia article on the Indian IT Act, 2000.
  2. Your savings-account balance page after you log in to net banking.
  3. The login page of that bank (the page itself, not the account).
  4. A doctoral thesis inside a university's online repository, reachable only through the university's own search box.
  5. The BBC's news mirror at an address ending in .onion.
  6. An investigative feature behind a news site's subscription wall.
  7. A company's internal wiki, on a server reachable only from its offices.
  8. The results page after you search for a product inside a shopping app.
  9. The case-status page for a court case, fetched from the court's portal by case number.
  10. A hidden market for stolen cards on Tor (classification only β€” you are not visiting it).
  11. Your email inbox.
  12. A public tweet linking a news story.

Expected result β€” answer key

1.  Surface β€” indexed, linked, public.
2.  Deep β€” behind login, generated on demand.
3.  Surface β€” the login page itself is public and indexed. The trap:
    the page that GUARDS the deep web can be surface web.
4.  Deep β€” exists only as a query result inside the repository's
    own search; general engines do not index it.
5.  Dark β€” needs Tor; and legitimate. Dark does not mean criminal.
6.  Deep β€” paywall.
7.  Deep β€” intranet: no public links, access-controlled.
8.  Deep β€” query-generated results inside an application.
9.  Deep β€” public record, but query-generated on an official portal.
10. Dark β€” onion service; criminal example, classified from
    description without visiting.
11. Deep β€” login; visible also to: the mail provider, and to anyone
    who compromises the account (password reuse, no MFA).
12. Surface β€” public and indexed; visible also to: the platform,
    its analytics partners, and archives that snapshot public pages.

Lab β€” Technical tier: install Tor Browser, visit legitimate onion services

⚠ Authorized practice only β€” the hands-on tier runs strictly inside your own isolated lab VMs (Chapter 5) or a legal practice platform. Never a real third-party system. Legal & ethical use.

Switch to Technical mode for the hands-on lab.

Objective: reach the dark web the way a professional does β€” through the official tool, with the download verified, visiting only legitimate onion services. This lab is safe and legal. It is about understanding the technology, not exploring markets.

Steps

  1. Get Tor Browser from the source. Download only from the Tor Project's own site: torproject.org/download.[16] Never from a mirror, an app-store clone or a link someone sent you β€” tampered copies repackaged with malware are a documented trick aimed at people seeking Tor.
  2. Verify the download. Follow the signature-verification instructions linked from the Tor Project's download and support pages.[17] Verification proves the file you received is the file they published. On first run, recent Tor Browser releases also verify themselves automatically β€” but knowing how to check a signature by hand is a skill every later chapter reuses.
  3. Connect. Open Tor Browser and press Connect. The startup screen shows the network being joined.
  4. Visit a legitimate search engine's onion. DuckDuckGo β€” the privacy-focused search engine that Tor Browser lists among its default search options β€” runs an official onion service. Its address begins duckduckgogg42xjoc72x3sjasowoarfbgcmvfimaftt6twagswzczad.onion; confirm it from Tor Browser's own search-settings page rather than trusting any third-party list. Unverified: check source (whether it appears as a listed default engine can vary by Tor Browser version and region β€” check your own settings page.)
  5. Visit a legitimate news onion. The BBC announced its Tor news mirror in October 2019; the announcement page on bbc.com carries the official address, which begins bbcnewsd73hkzno2ini43t4gblxvycyac5aw4gnv7t2rccijh7745uqd.onion.[13] Get the address from the BBC's own page over ordinary HTTPS, then open it in Tor Browser. Compare: same journalism as bbc.com, different transport. If the address fails to load, re-check it against the BBC's page β€” onion services do move, and stale addresses are a classic phishing vector (Chapter 21).
  6. Read the circuit. While on an onion site, open the circuit display (the site-information icon next to the address bar). You should see three relays and no exit relay β€” onion traffic never leaves the Tor network.
  7. Observe the address format. Modern addresses are 56 characters long β€” these are version-3 onion services. The old 16-character version-2 addresses stopped working entirely when Tor retired them in 2021.[18]
  8. Write three lines of reflection: what layer were you in; what was legitimate about every service you visited; and what you did not do (visit markets, click link-lists, download files). That discipline is the researcher's default stance in Chapter 27.

Expected output

- Tor Browser connects; the connection screen completes without bridges.
- The DuckDuckGo onion loads a familiar search page over its .onion address.
- The BBC onion loads BBC News content; the page identifies itself as the
  Tor edition.
- Circuit display for the onion site: "This site is an onion service",
  three relays listed, no exit node.
- Address bar shows a 56-character .onion address on both sites.

Quiz

Twenty questions: fifteen core and five marked Technical. Pass mark 60 per cent. Every answer carries an explanation.

Further reading

  • Bergman's 2001 paper β€” read the original and see how a careful database study became a legend[3].
  • The Tor Project's own history page β€” the primary narrative for the NRL-to-nonprofit story[9].
  • Christin's Silk Road measurement β€” the model for studying hidden services with numbers instead of adjectives[4].
  • Tor Metrics β€” spend ten minutes with the users and onion-services charts and their method notes[5].
  • The IMDEA Networks 2026 press release on the IEEE TIFS 2025 longitudinal onion-services study[6].
  • Bartlett's The Dark Net and Gehl's Weaving the Dark Web for the folklore and the legitimacy questions[8][14].
  • The BBC's own Tor-mirror announcement β€” an organisation explaining its onion service in its own words[13].

References

  1. US Department of Justice, US Attorney's Office, Southern District of New York, announcement of the indictment of Ross Ulbricht (arrested San Francisco, 1 October 2013), press release, 4 February 2014 β€” justice.gov.
  2. NPR, "Trump pardons Ross Ulbricht, creator of Silk Road", 21 January 2025 β€” npr.org.
  3. Bergman, M. K., "The Deep Web: Surfacing Hidden Value", Journal of Electronic Publishing 7(1), August 2001, University of Michigan β€” doi.org/10.3998/3336451.0007.104 (site blocks some automated checkers; opens in a browser).
  4. Christin, N., "Traveling the Silk Road: A Measurement Analysis of a Large Anonymous Online Marketplace", WWW '13, 2013 β€” andrew.cmu.edu (PDF).
  5. The Tor Project, Tor Metrics β€” users by country and unique onion services seen β€” metrics.torproject.org (figures read 16 October 2026).
  6. IMDEA Networks, "Spanish researchers conduct the largest ever survey of the Tor network…" (on RodrΓ­guez Barredo-Valenzuela, Pastrana and SuΓ‘rez-Tangil, "Snorkeling in Dark Waters", IEEE Transactions on Information Forensics and Security 20, 2025), press release, 29 June 2026 β€” networks.imdea.org.
  7. Chaum, D., "Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms", Communications of the ACM 24(2), February 1981 β€” dl.acm.org.
  8. Bartlett, J., The Dark Net, William Heinemann, 2014.
  9. The Tor Project, "History" β€” torproject.org/about/history.
  10. Hyphanet (formerly Freenet) β€” project site β€” hyphanet.org; Freenet project history β€” freenet.org/about/history.
  11. Dingledine, R., Mathewson, N. and Syverson, P., "Tor: The Second-Generation Onion Router", 13th USENIX Security Symposium, 2004 β€” usenix.org.
  12. I2P project, "20 Years of Privacy: A Brief History of I2P", 28 August 2021 β€” i2p.net.
  13. BBC News, "BBC News launches 'dark web' Tor mirror", 23 October 2019 β€” bbc.com.
  14. Gehl, R. W., Weaving the Dark Web: Legitimacy on Freenet, Tor, and I2P, MIT Press, 2018.
  15. India Code, Information Technology Act, 2000 (as amended) β€” indiacode.nic.in.
  16. The Tor Project, Tor Browser download page β€” torproject.org/download.
  17. The Tor Project, support portal (download verification guidance) β€” support.torproject.org.
  18. The Tor Project, "Onion Service version 2 deprecation timeline", blog, 2 July 2020 β€” blog.torproject.org.

Unverified: check source Live figures from Tor Metrics are estimates that move day to day; re-read the charts before quoting them in any report. The DuckDuckGo onion's presence among Tor Browser's listed default search engines may vary by release and region β€” verify in your own browser's settings. Figures in this chapter are current as of 19 October 2026 unless noted otherwise.