"The technique, scheduled to be demonstrated Thursday at the Black Hat security conference in Las Vegas, decodes encrypted data that online banks and e-commerce sites send in responses that are protected by the widely used transport layer security (TLS) and secure sockets layer (SSL) protocols. The attack can extract specific pieces of data, such as social security numbers, e-mail addresses, certain types of security tokens, and password-reset links. It works against all versions of TLS and SSL regardless of the encryption algorithm or cipher that's used."
"It requires that the attacker have the ability to passively monitor the traffic traveling between the end user and website. The attack also requires the attacker to force the victim to visit a malicious link. This can be done by injecting an iframe tag in a website the victim normally visits or, alternatively, by tricking the victim into viewing an e-mail with hidden images that automatically download and generate HTTP requests. The malicious link causes the victim's computer to make multiple requests to the HTTPS server that's being targeted. These requests are used to make "probing guesses" that will be explained shortly."
"It's the latest attack to chip away at the HTTPS encryption scheme, which forms the cornerstone of virtually all security involving the Web, e-mail, and other Internet services. It joins a pantheon of other hacks introduced over the past few years that bear names such as CRIME, BEAST, Lucky 13, and SSLStrip. While none of the attacks have completely undermined the security afforded by HTTPS, they highlight the fragility of the two-decade-old SSL and TLS protocols. The latest attack has been dubbed BREACH, short for Browser Reconnaissance and Exfiltration via Adaptive Compression of Hypertext."