Riiven Threads
HTTPS
The padlock on your browser is one court ruling, one bug, and one nonprofit away from disappearing.
You typed your card number into a website yesterday and did not think twice. The reason is a small padlock, and it is younger than it looks. The idea it rests on arrived in 1976. Netscape turned that idea into a handshake in 1995. A court had to rule in 1996 before the strong version was legal to send abroad. The cipher that actually carries your bytes was picked by open competition and published in 2001. And the certificate vouching that the site is who it claims cost money, until a nonprofit started giving them away. For most of the web's life the padlock was an upgrade you paid for. The question is what had to happen before it became the default.
- 1round trip
- What a modern TLS handshake costs before your data starts moving.
- 40bit keys
- The strongest encryption US law let you export through the 1990s. Breakable in an afternoon.
- 85%
- Web traffic on HTTPS once certificates became free, up from about 30% in 2014.
How each idea was handed down
HTTPS inherited its parts in sequence. Each field handed its result to the next.
The inheritance, in order
- 1976 Public-Key Cryptographyhanded down ↓
- 1995 SSL/TLS Protocol Designhanded down ↓
- 1996 Crypto Wars & Export Policyhanded down ↓
- 2001 Symmetric Cipher Standardizationhanded down ↓
- 2016 PKI & Certificate Authorities
Follow the inheritance, link by link.
Each link hands its result to the one after it. The chart above places them by date.
Keystone
Agreeing a secret in front of everyone
Two strangers can agree on a secret out loud, in front of everyone. Until 1976, this was considered impossible.
Diffie and Hellman, in 1976, showed that two people who have never met can shout numbers at each other in public and end up sharing a secret nobody listening can compute. That is exactly what your browser does in the first milliseconds of every HTTPS connection, with a server it has never contacted before. Without this one move, a shop would have to post you a key before you could buy anything, and internet commerce among strangers would be a logistics problem rather than a click.
Without this link
With only symmetric cryptography, a browser could not open an encrypted session with a server it had never contacted before. Every site would have to distribute keys to every visitor in advance, by post or in person, which is not a slow web but no web commerce at all. The pairwise key count for n parties grows as n(n-1)/2, so a billion users need about 500 quadrillion keys exchanged before the first page loads.
Without public-key cryptography, a billion users would need about 500 quadrillion pre-shared keys between them, more than the seconds since the Big Bang.
How we know
Diffie & Hellman's 1976 paper New Directions in Cryptography proposed that two parties could establish a shared secret over a public channel without prior key exchange, a possibility many cryptographers had argued was logically impossible. RSA (Rivest, Shamir, Adleman, 1978) provided the first concrete instantiation. This single conceptual move made internet-scale encryption between strangers feasible.
Source: New Directions in Cryptography (1976) · tier1
The idea existed by 1976. Turning it into something a browser could actually perform took another twenty years.
Wrapping the math into a conversation
Knowing two strangers can share a secret is not a protocol. Someone had to decide who speaks first.
A handshake has to settle everything before a single byte of your card number moves: which cipher both sides know, which certificate proves the server is who it claims, and which fresh key this one session will use. Taher Elgamal's team at Netscape wrote that choreography into SSL in the mid-1990s. The version you actually use is its descendant, standardized as TLS in 1999 and rewritten in 2018 to cut the handshake to a single round trip. Get the order of those steps wrong and the strongest cipher in the world protects nothing.
Without this link
Without an agreed handshake, every site and browser would invent its own way to negotiate keys and prove identity. Nothing would interoperate, and the negotiation step, which is where nearly every real-world TLS attack has landed, would have no common definition to audit or fix.
TLS 1.3 opens an encrypted session in a single round trip. The handshake is the part attackers target, and the part that had to be standardized before anything else could be.
How we know
TLS 1.0 (RFC 2246, 1999) split the protocol into a handshake layer that negotiates parameters and a record layer that carries encrypted application data. Every later attack of consequence, from BEAST to FREAK to Logjam, targeted the negotiation rather than the ciphers themselves, which is why TLS 1.3 in 2018 deleted the downgrade paths outright.
Source: RFC 2246: The TLS Protocol Version 1.0 (1999) · tier1
The handshake could agree on a key. Whether that key was legal to use at full strength abroad was still an open question.
The export rule that kept encryption deliberately weak
Until 1996, U.S. law treated strong encryption as a weapon. The padlock you trust today was, technically, a munitions violation.
Through the 1990s, exporting software with keys stronger than 40 bits was illegal. Forty bits was weak enough for a graduate student to break in an afternoon. A federal court ruled in 1996 that source code is protected speech, which broke the legal foundation of the export rules. The math had been ready for two decades. The law had to catch up before HTTPS at modern strength was allowed to leave the country.
Without this link
Export-grade SSL used 40-bit RC4, brute-forceable in about three hours on commodity 2015 hardware. The FREAK attack in 2015 showed those weakened code paths were still reachable in modern browsers and could be triggered to downgrade a live HTTPS connection, affecting an estimated 36% of HTTPS sites. A rule written in the 1990s was still breaking the padlock twenty years after it lapsed.
How we know
Until the late 1990s, U.S. law classified strong cryptography as a munition; software using keys longer than 40 bits could not legally be exported. The Bernstein v. United States ruling (1996) found that source code is protected speech, and successive court and administrative decisions liberalized export controls by 2000. Without this legal turn, browsers shipped abroad were limited to crypto a research student could break in an afternoon.
Source: Bernstein v. United States Department of Justice (1996) · tier1
Once strong encryption could leave the country legally, it still cost money every site had to find.
The cipher that carries the actual bytes
Public-key math is far too slow to encrypt a video. It is used once, to agree a key, and then it steps aside.
The padlock does its real work with a symmetric cipher, where both ends hold the same key. That is thousands of times faster than public-key math, which is why the handshake uses the slow method once and the fast method for everything after. Two Belgian cryptographers, Joan Daemen and Vincent Rijmen, won the open public competition to pick that cipher, and NIST published their design as AES in 2001. Choosing it in the open, rather than behind a government door, is why the world was willing to trust it.
Without this link
Without a standard, openly vetted bulk cipher, every implementation would pick its own, and browsers and servers would have no fast algorithm they could both be sure of. Public-key operations alone are orders of magnitude too slow to encrypt real traffic, so the connection would be either unusably slow or protected by something nobody had reviewed.
Fifteen ciphers were submitted to an open competition and attacked in public before one became AES. That openness is what made it trustworthy.
How we know
NIST ran the AES process from 1997 to 2000 as a public competition: fifteen submissions, open cryptanalysis by anyone who wanted to attack them, and three international conferences before Rijndael was selected. FIPS 197 followed in 2001. The contrast with the closed selection of DES in the 1970s, which left decades of suspicion about the S-box design, is the whole argument for doing it in public.
Source: AES development, the open competition that selected Rijndael (2001) · tier1
Fast, trusted encryption still leaves the last question: the browser has a key, but whose key is it?
The price of trust fell, and the padlock spread
The math proves you have a key. It does not prove the key is yours. Someone has to vouch for that, and that someone has to be paid.
When your browser shows a padlock, it is not trusting the website. It is trusting a third party, a Certificate Authority, that signed the site's key after checking who they were. For twenty years that signature cost money, which is why HTTPS sat at about 30% of the web in 2014. Let's Encrypt, founded in 2014, began issuing them free and automatically, and adoption passed 85%. The padlock spread because the price of trust fell to zero.
Without this link
Without PKI, RSA gives you encryption with someone, but no way to verify who. Man-in-the-middle attacks become trivial. The 1990s web of trust (PGP) failed to scale to consumers; centralized paid CAs got HTTPS to roughly 30% of page loads by 2014. Let's Encrypt (free, automated, ACME protocol, 2016) collapsed the cost barrier and pushed adoption past 85% by 2023.
HTTPS adoption: ~30% of page loads in 2014 (paid CAs only) → ~85% by 2023 after Let's Encrypt made certificates free
How we know
Public-key cryptography proves you hold a private key, but says nothing about whose key it is. Certificate Authorities (Verisign from 1995, then dozens more, then Let's Encrypt) sign public keys after verifying identity, and your browser ships with a list of the ones it trusts. Who runs that ecosystem, who pays for it, and who is liable when one of them signs the wrong thing are economic questions the math cannot answer.
Source: Let's Encrypt: An Automated Certificate Authority to Encrypt the Entire Web (2019) · tier1
Watch
A visual companion to the fields above.
Takeaway
The padlock is less a feature than a settlement, agreed piece by piece by cryptographers, a standards committee, a federal judge, and eventually an accountant. Diffie and Hellman said in 1976 that strangers could agree a secret in the open. Netscape wrote the choreography for doing it in a browser. A 1996 ruling that source code is speech is the reason the strong version could leave the country at all. An open competition, not a government committee, chose the cipher. And the last barrier was not mathematical or legal but a line item: certificates cost money, so most of the web went without until Let's Encrypt made them free. By 2018 Chrome had inverted the default, marking unencrypted pages as the suspicious ones. The padlock stopped being a feature and became the floor. What is worth remembering is how recent that is, and that the last thing standing between you and it was a bill nobody wanted to pay.
References
- New Directions in Cryptography (1976) tier1
Diffie & Hellman, IEEE Transactions on Information Theory vol. IT-22 (1976). The paper that proposed public-key cryptography as a category and showed key exchange over an insecure channel was possible.
- RFC 2246: The TLS Protocol Version 1.0 (1999) tier1
Tim Dierks & Christopher Allen, IETF, January 1999. The first standards-track version of the protocol Netscape shipped as SSL, defining the handshake and record layers HTTPS still runs on.
- Bernstein v. United States Department of Justice (1996) tier1
Northern District of California, 1996; affirmed by Ninth Circuit, 1999. Ruled that cryptographic source code is protected speech under the First Amendment, breaking the legal foundation of the export-control regime.
- AES development, the open competition that selected Rijndael (2001) tier1
National Institute of Standards and Technology, archived project record. Fifteen candidate algorithms were announced at the First AES Candidate Conference in 1998, analyzed in public, narrowed to Rijndael in 2000, and published as FIPS 197 in 2001.
- Let's Encrypt: An Automated Certificate Authority to Encrypt the Entire Web (2019) tier1
Aas, Barnes, Case, Durumeric, Eckersley, Flores-López, Halderman, Hoffman-Andrews, Kasten, Rescorla, Schoen & Warren, ACM CCS 2019. The reference paper on the certificate authority that broke the cost barrier of HTTPS deployment.
