The defense proposed by Lidar and Wu is to give each legitimate node on the network a list of times distributed over a long period, during which the quantum computers can exchange information and possibly entangle qubits across the network. All other times, a set of decoy qubits are sending data on the network to mimic legitimate traffic flow. When not connected, the networked qubits transfer data back to a third set of qubits, known as the ancillary, or ancilla, qubits. Those registers are never connected to the network, so quantum computations can be carried out without worry of attack.
"As long as we can keep the local nodes free from malicious intruders and build a heavily fortified castle around them, we can assume the ancilla qubits are malware free," Lidar said.
Lidar and Wu found in a paper published earlier this year (PDF) that the number of network connections can never exceed the ratio of the average time between attacks over the length of time it takes to complete an attack.
In classical computing the result would be a poor tradeoff. Only connecting for one second out of every hundred or thousand would unacceptably slow down the calculation speed of, say, a grid computing system. However, in a quantum computing system, such a slowdown does not appreciably affect the speed gains of the system.
"The protocol shuts down the network for a certain period of time, say 99 percent," Lidar said. "While that's a factor of 100, it doesn't matter, because it doesn't change the complexity of the problem."
In other words, for a computer that exists in multiple dimensions and uses a form of teleportation in its calculations, taking a hundred, or even a thousand, times longer means little.