What is the HTTPS bicycle attack?

The HTTPS bicycle attack is an attack in which the attacker can capture the HTTPS traffic of a user and exploit the TLS packets to deduce sensitive information like the length of a password, GPS location, or IPv4 address range of the user. The attack was first discovered by security researcher Guido Vranken and released on December 30, 2015.

How can an attacker deduce sensitive information like password length from the captured HTTPS traffic?

When sensitive data is transferred between two hosts over an unsecured network, it often uses a TLS connection. The sensitive data sent by the user is encrypted using a secret key and then transferred to the server. Usually, decrypting the sensitive data without the secret key is extremely difficult.

But, for HTTPS traffic, the plaintext HTTP headers included in every packet can be exploited by the attacker. The attacker can use the side-channel information (What is the side-channel information?) and deduce the length of particular components transferred in that particular request.

At this point, if the attacker already has prior information on the length of some of the components out of the combined components, he can subtract the length of the known components from the length of combined components and deduce the length of the sensitive data.

For example, suppose the attacker wants to deduce the length of a user’s password for a particular target website. The attacker can capture the encrypted TLS packets sent during the authentication requests. At the same time, he may collect information on the browser the user is using. The attacker can do it easily by perpetrating a direct attack on the user. An insecure HTTP request easily reveals the user-agent string that contains information about the user’s browser.

Now, the attacker can replicate browser requests to the target website using the same browser and deduce the length of the requests to various pages on the target site. And, from the encrypted TLS payload of the browser requests, he can extract the length of the payloads.

The attacker can now compare the Pearson correlation coefficient for the plain texts and encrypted requests. On comparing the results, he may deduce which encrypted request is for which page on the website.

Now, the attacker has information on which encrypted TLS request is for the login page. From this, the attacker can subtract the length of known headers the user’s browser has sent. He can also deduce the length of possible cookies that are static enough in length.

Now, for an authentication request, normally, the username and passwords are sent together. At this …