Multipath TCP aims to allow a TCP connection to use multiple paths to maximize resource use and increase redundancy, and enables the simultaneous use of several IP addresses and interfaces.  The result looks just like a regular TCP interface to applications while actually spreading data over several subflows behind the scenes.  This results in better resource utilization, better throughput, and better reaction to failures.

In short, it creates more robust and dependable connections and vastly improves performance, especially for delay-sensitive applications.  It does this by breaking resources into multiple lower data rate clinks, so in addition to the throughput gains, links may be added or dropped as the user moves in or out of coverage without disrupting the TCP connecting.  Handoffs and failover are made easier by abstraction in the transport layer without any special mechanisms at the network or link level.

It’s especially good for wireless networks.  Being fully backwards-compatible with regular TCP, multipath extensions on top of regular TCP provides tons of benefits for mobile networks, like easy roaming and the ability to aggregate links for higher availability.

MPTCP is part of an ongoing open-source project by the Internet Engineering Task Force, intended to progress the internet’s transformation from a text-and-images network to one that supports increasingly demanding types of content and larger files.

Despite all the benefits it offers for a wide variety of network types, it also presents a potential security risk in the form of cross-path traffic fragmentation, especially since many network security technologies aren’t fully prepared.  For example, IPS systems can’t properly inspect them, presenting a potential way for attackers to sneak in.

However, that hasn’t stopped it from being a popular choice even among major networks.  In 2013, Apple made news by using the then-new protocol to link iOS devices to their cloud-based voice command service Siri.