What Is SSL And How Does It Work?

SSL/TLS, Cryptography, SSL Certificates, and the Infamous Heartbleed Vulnerability

Steven Li
10 min readAug 30, 2017
Browsers let users know when they are communicating with an SSL-enabled web-server with a padlock icon, assuring them of the authenticity of the server and that their communication will be private.

What is SSL and how does it work?

SSL (Secured Sockets Layer) is a web protocol developed by Netscape in the 90s for enhancing web security. TLS (Transport Layer Security) was developed by the Internet Engineering Task Force (IETF) as an improvement on SSL. Generally, SSL is used to refer to both SSL and TLS.

SSL allows web clients and servers to:

  1. Verify each other’s identity.
  2. Encrypt messages they send to each other.
  3. Ensure the integrity of messages sent between them.

It works through a combination of symmetric and asymmetric encryption using cryptographic keys tied to an SSL certificate.

What is Symmetric Encryption and Asymmetric Encryption?

Encryption can be generally classified as two types: symmetric and asymmetric (also known as public key encryption).

Symmetric encryption means the ability to encrypt a message automatically confers the ability to decrypt. This is because it encrypts and decrypts using the same secret key. For example, an ancient symmetric encryption scheme is the Caesar Cipher. Modern symmetric encryption schemes include: Data Encryption Standard (DES), and Advanced Encryption Standard (AES).

Asymmetric encryption, aka public key encryption, is an encryption scheme that uses a public key and a private key. The public key is used for encryption while the private key is used for decryption. Unlike symmetric encryption, asymmetric encrypting does not automatically confer the ability to decrypt. Modern asymmetric encryption schemes include RSA, which relies on the difficulty of factoring large primes and Diffie-Hellman, which relies on the difficulty of solving discrete logarithms.

Asymmetric encryption can also be used to generate digital signatures. Much like their analog counterparts, digital signatures verify that a party approves of a document. Generating the signature is done with the private key, while verifying the authenticity of the signature is done using the public key. SSL uses these digital signatures to authenticate the identity of web servers.

Here’s a more in-depth article on the mathematics behind these encryption schemes.

What is an SSL certificate?

An SSL certificate is a digital document tied to a specific subject, such as a server hosted on a web domain. Its contents consists of:

  1. Who the SSL certificate is issued to (aka the subject).
  2. The expiration date of the SSL certificate
  3. The SSL certificate’s public key, which will be used for encryption.
  4. The digital signature of the certificate authority (more on this later) who issued the certificate.

The SSL protocol uses SSL certificates to

  1. Verify the authenticity of the server.
  2. Encrypt messages sent between client and server.
  3. Check the integrity of the messages sent between client and server.
Using an open source SSL library, we can inspect the public key of the SSL certificate issued to google.com. Note the SSL protocol uses IP port 443. Also note the x509 argument specifies we want to view this key in the x509 standard for public keys.

Where do SSL certificates come from?

Except for self-signed certificates, SSL certificates are signed and issued by certificate authorities, third party corporations who verify the identity of organizations in exchange for the certificate.

As noted above, there are self-signed certificates, which are not issued by certificate authorities. These are typically used for development purposes by web application developers.

The three largest certificate authorities, ranked by certificates issued, are Symantec, GoDaddy, and Comodo. These three organizations issued more than 75% of all SSL certificates, but there are hundreds of other certificate authorities.

What are self-signed SSL certificates? What other types of SSL certificates are there?

Self-signed certificates are not used for authentication since they aren’t issued by a certificate authority. But they can still be used for encryption. These certificates will trigger the browser is raise a warning for the user. Self-signed certificates are typically used by web development teams as a cheap solution to setting up SSL-enabled web servers for testing/development.

Self-signed certificates will cause the browser to raise a warning. These can be ignored as long as you are not sending sensitive information like credit card numbers or passwords to the server, and don’t care if the web server sends back false information.

But even among SSL certificates issued by certificate authorities, there are three types of certificates, each with varying levels of security (and prices):

  • Domain Validated. These account for 70% of all certificates. They are the cheapest certificates with the lowest level security as they only verify that the domain associated with the certificate is registered and owned by the person who requested the certificate. These do not verify that the person is affiliated with a legitimate company.
  • Organization Validated. These certificates are more expensive and have a higher level of security than domain validated. The organization requesting the certificate needs to prove to the certificate authority that it is a valid company and it registered and owns the domain that the certificate is associated with.
Chrome uses a padlock to inform the user the website with domain validated or organization validated SSL certificates.
  • Extended Validation. These certificates are top of the line. These are the most expensive certificates and have the highest level of security. As such, the certificate authority will require the company to go through a rigorous vetting process. For example, all details about the company must be verified by the certificate authority.
Chrome uses a padlock but also the organization’s name in green to inform the user the website is using an extended validation SSL certificate.

How does the client know an SSL certificate has been issued by a legitimate certificate authority?

Clients like browsers or smart phones, will come with a list of some certificate authorities and their public keys. However, there are hundreds of certificate authorities, so the client can’t store them all. Instead, the client only needs to know a few root certificate authorities.

These root certificate authorities can issue special certificates to intermediate certificate authorities (after vetting them of course), and these intermediate certificate authorities can issue SSL certificates to everyone else.

For example, if a browser receives an SSL certificate that is signed by a root certificate authority, the client can treat the certificate as valid. But if the client receives a certificate that is signed by an intermediate certificate authority that it doesn’t know, that certificate will include the signature of the root certificate authority that issued the certificate to the intermediate certificate authority. The client knows this root certificate authority so it can treat the certificate as valid even though it was not issued by them directly.

From the chrome developer tools’s security tab, we see that foxpass.com uses an SSL certificate issued by RapidSSL, an intermediate certificate authority, who was in turn issued a certificate by Geotrust, a root certificate authority. We call this a certificate chain.

So how does the SSL protocol actually work?

SSL is actually built on top of the TCP layer, so after a TCP connection is established, the client and server engage in what is called the SSL handshake.

The client will send the server the following:

  • Which version of SSL (there are several) it is running.
  • Which cipher-suites it supports.
  • Which compression methods it supports.

The server will select the highest version of SSL it supports, the cipher-suites, and compression methods.

The server will offer an SSL certificate to confirm its identity. The server may also request that the client authenticate itself as well. The client will attempt to verify the server’s certificate. After verifying the certificate by referencing the certificate chain against the client’s list of root certificate authorities, the client is now confident on the server’s identity. More importantly, the client is confident about the authenticity of its public key.

A random secret key will be generated by the client. The client will then encrypt this random secret key through asymmetric encryption using the server’s public key and send it to the server. Because the server has the corresponding private key, it can decrypt the client’s random secret key.

Both client and server will now have this random secret key. The SSL handshake is over and the two applications can communicate securely by symmetrically encrypting messages with this secret random key.

This random secret key is also used by the Message Authentication Code (MAC) algorithm which generates an authenticity token that accompanies each message sent between client and server. This MAC token is used to verify that the message has not been modified. That is, if a hacker modified a message between client and server without knowing the secret random key, she won’t be able to generate the correct MAC token for the message. As such, the server will know the message has been tampered with.

Diagram from RFC 6101 illustrating the SSL handshake. Note that ClientHello is the first message that the client sends to server to initiate the SSL handshake. So polite!

How do we install an SSL certificate on a server?

Each certificate authority and web server (ie. Apache, Nginx) will have slightly varying instructions, but the general steps are the same.

First install the OpenSSL library (an open source implementation of SSL) on the server that needs to serve up the certificate. Note some versions of OpenSSL have a serious security flaw called the Heartbleed vulnerability (more on this later). The versions are:

  • OpenSSL 1.0.1–1.0.1f (inclusive) are VULNERABLE
  • OpenSSL 1.0.1g is NOT VULNERABLE
  • OpenSSL 1.0.0 is NOT VULNERABLE
  • OpenSSL 0.9.8 is NOT VULNERABLE

After installing a secure version of OpenSSL, use it to create a certificate signing request (CSR). This CSR will generate a .key file which will be the private key associated with the certificate as well as .crs file which will be copied into a certificate authority’s certificate application form.

When the certificate authority approves of the request and returns a certificate .crt file, place it and the private key .key into a directory that looks something like /etc/<web_server>/ssl depending on the web server. There will also be some configuration files that need tweaking. For example, nginx requires the creation of a /etc/nginx/snippets/ssl-certs.conf file and a /etc/nginx/snippets/ssl-params.conf file which will specify which SSL versions and cipher-suites to use, etc. The /etc/nginx/sites-available/default file’s server block will need the declarative:

listen 443 ssl http2 default_server; 
listen [::]:443 ssl http2 default_server;
include snippets/ssl-certs.conf;
include snippets/ssl-params.conf;

What is SSL Termination?

High traffic web applications are powered by multiple upstream web servers behind a load balancer. In these cases, the SSL certificate does not need to be installed on every one of the upstream web servers. Instead, it is installed on just the load balancer. This is called SSL termination as the load balancer decrypts incoming requests and encrypts outgoing responses so the upstream web servers don’t have to worry about SSL.

What was the SSL Heartbleed Vulnerability?

An SSL heartbeat is when the client or server sends the other a “keep-alive” message to check if the other party is still listening and still using the same encryption/decryption parameters. By maintaining these session parameters, they don’t have to periodically renegotiate security parameters.

The heartbeat check consists of sending a message along with the number of bytes (up to 64Kb) of that message to the other party. The other party simply echoes the message back.

In some versions of OpenSSL, the heartbeat code had a bug, introduced in 2012, that did not check that the message length is equal to the claimed number of bytes of the message. As a consequence, if an application is running a vulnerable version of OpenSSL, a hacker can send a heartbeat with a message of only 1 byte but claim it was 64kB. The application would send back 64kB minus 1 byte of its memory.

Note that most of the time, the hacker does not get useful information from this; just random useless data from memory. But if a hacker keeps trying enough times, she may get something useful like a password or a cookie.

It’s important to emphasize the heartbeat bug was not due to a flaw in the TLS protocol, but a bug in the OpenSSL library implementation of TLS. This bug was disclosed in 2014 and patched.

As per usual, XKCD explains the heartbleed bug with an easy-to-understand comic.

Aside from Web Browsers and Web servers what else uses SSL?

We’ve been discussing SSL primarily in the context of securing web browser and server communication with HTTPS, the SSL-enabled version of HTTP. But it’s used in many other protocols.

SSL can be used to secure the file-transfer protocol (FTPS), and the email protocol (SMTPS). The Windows Update application uses SSL to ensure the authenticity of Windows patches. A VPN can be built on top of SSL. SSL can be used to secure connections between MYSQL clients and servers.

But in the most general sense, SSL is built on top of TCP, meaning any TCP/IP based protocol can incorporate SSL into it if needed.

What are all the versions of SSL/TLS?

  • SSL version 1.0 was developed by Netscape in the early 90s but never released due security issues.
  • SSL version 2.0 was released by Netscape in 1995, but still had many security flaws.
  • SSL version 3.0 was released by Netscape in 1996, to address the security flaws in version 2.0. This version varied significantly from previous versions. However, as of 2015, SSL 3.0 has been deprecated.
  • TLS version 1.0 was released in 1999, and based significantly on SSL version 3.0 with minor security improvements. TLS version 1.0 is essentially SSL version 3.1.
  • TLS version 1.1 was released in 2006, with significant changes over 1.0.
  • TLS version 1.2 was released in 2008, with significant changes over 1.1.
  • As of July 2017, TLS 1.3 is a working draft, and contains significant improvements over 1.2.

What cipher-suites are used by SSL?

For key exchange and authentication algorithms:

  • Rivest, Shamir, Adleman (RSA)
  • Diffie-Hellman (DH)
  • Diffie-Hellman Ephemeral (DHE)
  • Elliptic-Curve Diffie-Hellman (ECDH)
  • Kerberos (KRB5)
  • Secure Remote Password Protocol (SRP)
  • Pre-shared Key (PSK)
  • DSA Digital Signature Algorithm
  • ECDSA Elliptic Curve Digital Signature Algorithm
  • DSS Digital Signature Standard

For symmetric encryption and MAC algorithms:

  • 3DES Triple Data Encryption Algorithm
  • AES Advanced Encryption Standard
  • Camelia Block cipher developed by Mitsubishi and NTT
  • DES Data Encryption Standard
  • Fortezza Security token based cipher
  • GOST Block cipher developed in USSR
  • IDEA International Data Encryption Algorithm
  • RC2 Rivest Cipher 4
  • RC4 Rivest Cipher 2
  • SEED Block cipher developed by Korean Information Security Agency SHA
  • Secure Hash Algorithm
  • MD5 Message Digest algorithm 5

If there are any errors, please leave a comment below. Thanks — S

--

--

Steven Li
Steven Li

Written by Steven Li

Writing About Rails, React, Web Application Technology, Databases, and Software Engineering

Responses (5)