Instead of plastic domes, they use a sensor network to work out where potential eavesdroppers are, and speakers to generate a subtle masking sound at just the right level.
It sounds simple, but it needs quite a bit of infrastructure. The walls of the room must be peppered with light-switch-sized units that include a microphone, a speaker, an infrared location sensor and networking circuitry connected to a server. When somebody wants to activate what the MIT researchers call the “sound shield”, they do so on their desktop computer. Knowing the position of the computer, the sensors identify the person and map out the locations of people around them. Software assesses who is so close that they must be participants in the conversation, and who might be a potential eavesdropper.