Data delivered over an unencrypted channel (e.g. HTTP) is insecure, untrustworthy, and trivially intercepted. TLS can help!
TLS (also known as SSL) is the industry standard for providing communication security over the Internet.
TLS guarantees identification, confidentiality, and integrity between a client (a computer) and a server.
In other words, TLS ensures that a Man-in-the-Middle (MitM) can't snoop or tamper with an Internet connection between a user and website. A man-in-the-middle (MiTM) is a term used to describe a third party that can passively monitor and/or actively tamper with a connection between two unknowing parties. A MiTM attacker relays messages between two parties, making them believe that they are talking directly to each other, when in fact the entire conversation is controlled by the attacker.
MiTM attacks happen in real life! Here are some recent examples:
TLS only protects the connection between your computer and the server. It does not protect data on the client or data on the server. This means:
HSTS is a mechanism enabling web sites to declare themselves accessible only via secure connections and/or for users to be able to direct their user agent to interact with given sites only over secure connections. Chrome supports HSTS and comes preloaded with a set of domains that use HSTS by default. More details, including how to add a site to Chrome's preloaded HSTS list, here.
TLS relies on websites serving authenticated (X.509) certificates to prove their identities, which prevents an attacker from pretending to be the website. Certificates bind a public key and an identity (commonly a DNS name) together and are typically issued for a period of several years.
Chrome has HTTPS "pins" for most Google properties — i.e. certificate chains for Google properties must have a whitelisted public key, or it will result in a fatal error. This feature helped Google detect a widespread MITM attack to Gmail users in 2011. You can read more about pinning here. There's also an Internet-Draft for HTTP-based public key pinning.
Sometimes events occur that invalidate the binding of public key and name, and the certificate needs to be revoked. For example, a major flaw in the implementation of OpenSSL left site operators' private key vulnerable to theft, so operators needed to invalidate their certificates. Revocation is the process of invalidating a certificate before its expiry date. Chrome uses CRLSets to implement certificate revocation. You can read about the how and why of Chrome's certificate revocation in our Security FAQ.
If there is an error in the certificate, Chrome can’t distinguish between a configuration bug in the site and a real MiTM attack, so Chrome takes proactive steps to protect users.
If a site has elected to use HSTS, all certificate errors are fatal.
This means that the certificate chain for the current page is contains a certificate using a SHA-1-based signature, which is outdated and deprecated in Chrome. There are two criteria that determine which lock icon is shown in Chrome:
Starting in Chrome 42, the following logic applies:
You may see:
This means that the connection to the current website is using an outdated cipher suite (which Chrome still allows if the server insists on it).
In order for the message to indicate “modern cryptography”, the connection should use forward secrecy and either AES-GCM or CHACHA20_POLY1305. Other cipher suites are known to have weaknesses. Most servers will wish to negotiate TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256.
As of May 2015, Chrome allowed DHE with 512-bit groups. Based on new research, the minimum requirement is changing to 1024 bits in the medium-term. See this email thread and this page.
This actually means that the connection is using HMAC-SHA1 for data integrity, rather than as a certificate signing algorithm (e.g. sha1WithRSAEncryption). The HMAC construction is strong enough that it is not broken when used with SHA1 (or even MD5) as the hash function, so this is not currently deprecated.
When properly deployed, TLS provides three guarantees: authentication of the server, data integrity (tamper-evidence), and data confidentiality. People often think TLS and HTTPS only apply in threat scenarios where data confidentiality is needed, but in fact they apply when any (or, most often, all 3) guarantees are beneficial.
My site doesn't need TLS. I'm not a bank.More people are connected to the web than ever before and from more places and more devices (laptops, phones, tablets, and other things). Very often, this access is over untrusted or hostile networks. Data delivered over a clear text protocol, like HTTP, is insecure, untrustworthy, and trivially intercepted. Neither the user / user-agent nor the web server / application can trust that the data was not tampered with or snooped by some third party - that's a terrible situation for both users and web site operators!
With so much of people's lives moving online, it’s imperative developers take steps to protect their sites' and users' data, which can even include the mere usage of a web site. By analyzing and correlating the sites and pages a user visits, observers like schools, ISPs, and governments can learn quite a bit about a user that the user would wish to keep confidential, such as a users' sexual orientation (http://blogs.wsj.com/digits/2010/03/12/ftcs-privacy-worries-prompt-netflix-to-cancel-contest/) or physical location (http://www.theregister.co.uk/2009/05/21/geo_location_data/).
Historically, TLS used to have a significant performance on web applications. So, istlsfastyet.com? (Spoiler: Yes!) Check out https://istlsfastyet.com/ for more details and a performance checklist.
SSLs.com offers certificates for a very low price, as low as $5.
SSLmate.com is cheap and easy to use — you can buy certificates from the command line.
In Summer 2015, the Let’s Encrypt project will be offering free certificates.
For example, when used to secure HTTP traffic (i.e. HTTPS), we’re piggybacking HTTP entirely on top of TLS. This means the entirety of the HTTP protocol can be encrypted (request URL, query parameters, headers, and cookies), however, because host IP addresses and port numbers are necessarily part of the underlying TCP/IP protocols, a third party can still infer these. Also, while you can’t infer the contents of the communication, you can infer the amount and duration of the communication. For specific applications, it’s been demonstrated that this can leak useful information for an attacker, and services have added padding to counter the timing or pattern analysis.
The identity of the site you are visiting is still (unfortunately) pretty visible to passive eavesdroppers. For example, the IP addresses of client and server are shown in the clear on the network, and the hostname(s) of the sites you are visiting are transmitted in the clear in DNS requests, in the Server Name Indication portion of a TLS handshake, and in the server's certificate(s).
Also, since TLS is a transport protocol, attacks at other layers of the network stack remain. In particular, IP-level threats (e.g. spoofing, SYD floods, DDoS by data flood) are not protected and TLS doesn’t address common web application vulnerabilities, like cross-site scripting or cross-site request forgery.
SSL Labs puts out a great Deployment Best Practice Guide that should help site operators avoid the most common deployment mistakes. You can also test your setup via https://www.ssllabs.com/ssltest/.
To fix mixed content issues, make sure that all the resources loaded by an HTTPS page are also sent over HTTPS. If the resources are available on the same domain, you can use hostname-relative URLs (e.g. <img src="something.png">). You can also use scheme-relative URLs (e.g. <img src="//example.com/something.png">) — the browser will use the same scheme as the enclosing page to load these subresources. If the server does not serve these resources over HTTPS, you may have to serve them from elsewhere or enable HTTPS on that server.