GB2593697A

GB2593697A - Improvements to satellite-based QKD

Info

Publication number: GB2593697A
Application number: GB2004606.6A
Authority: GB
Inventors: lord Andrew
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 2020-03-30
Filing date: 2020-03-30
Publication date: 2021-10-06
Also published as: GB202004606D0

Abstract

Disclosed a computer-implemented method of optimising a quantum key distribution (QKD) system. The QKD system is adapted to perform a QKD session that involves the propagation of a plurality of photons from a first device to a second device. One of the first and second devices is located aboard a satellite in orbit. The method comprises running a machine learning algorithm whose inputs include values representing one or more of: (i) atmospheric conditions in the vicinity of a path from the first device to the second device; (ii) a spatial separation of the first device and the second device; or (iii) a measure of the performance of the first and/or the second device. The output of the algorithm is an estimate of the rate at which the first device and the second device generate bits of a quantum key.

Description

Improvements to satellite-based QKD QKD (Quantum Key Distribution) is a technique for distributing very secure symmetrical encryption keys between parties. It involves the transmission of a stream of photons between the parties. This places an upper limit on the spatial separation of the parties.

If the parties are located too far apart, attenuation of the photon stream becomes significant.

A solution to this difficulty is satellite-based QKD. In an example embodiment of this, a quantum transmitter aboard a satellite in low earth orbit, performs QKD with a quantum receiver on the ground and establishes a first quantum key. The satellite continues its orbit and performs a further QKD session with a second quantum receiver, located remotely from the first, establishing a second quantum key. The transmitter encrypts the first quantum key using the second and transmits it to the second receiver. The first and second receivers therefore each have the first quantum key. They can use this key to communicate with each other securely. Thus satellite-based QKD enables quantum-secured communication between remote parties.

A disadvantage of satellite-based QKD is that it involves transmitting the photon stream through the earth's atmosphere. This can cause problems. For example, it can be difficult to achieve directional alignment of the transmitter with the receiver. Also, attenuation of the photon stream can be significant.

It is desirable to overcome and/or mitigate the above-mentioned and/or other problems

associated with the prior art.

According to a first aspect of the invention there is provided a method of optimising a QKD system, the QKD system being adapted to perform a QKD session that involves the propagation of a plurality of photons from a first device to a second device, one of the first and second devices being located aboard a satellite in orbit, the method comprising running a machine learning algorithm, the inputs of the machine learning algorithm comprising one or more values of one or more of the following factors: (i) Atmospheric conditions in the vicinity of a path from the first device to the second device; (ii) The spatial separation of the first device and the second device; and (iii) A measure of the performance of the first and/or the second device, wherein the output of the algorithm is an estimate of the rate at which the pair generate bits of a quantum key.

An advantage of the invention is that it enables the estimation of the likely quantum key bit rate resulting from a QKD session between a satellite-based quantum communication device and multiple candidate ground-based quantum communication devices, considering the prevailing conditions. This enables the ground-based quantum communication devices to be ranked in order of likely productivity. The satellite can then perform QKD sessions with the ground-based devices in order of their ranking. This greatly increases the key generation rate of the system.

The first device may be a quantum transmitter and the second device may be a quantum receiver. Alternatively, the first device may be a quantum receiver and the second device may be a quantum transmitter. In presently preferred embodiments the first device may be the device located aboard the satellite and may be a quantum transmitter. In these embodiments the second device may be located on the earth and may be a quantum receiver.

The pair may be capable of performing QKD according to prepare-and-measure or quantum entanglement protocols.

The quantum communication device that is located on the earth may be located on land, and may be located in a building which may be a ground station. Alternatively the device may be located at sea, for example, aboard a ship.

The atmospheric conditions in the vicinity of a path from one device of the pair to the other may comprise air pressure, humidity, cloud coverage and pollution. Pollution may comprise light pollution and air pollution.

The method may comprise obtaining the one or more values for the one or more of the factors (i)-(iii). This may comprise measuring the one or more values for the one or more of the factors (i)-(iii) at the satellite and/or at the earth station, or receiving such measured value(s). Alternatively or in addition this may comprise obtaining the one or more values from existing recorded data, stored locally or remotely, which may be public data.

The one or more values for the spatial separation of the pair may be obtained from a global positioning measurement system such as GPS. The one or more values for a measure of the performance of one or both of the pair of devices may be a data transmission rate or may be a data detection rate. In embodiments in which factor (iii) is a measure of the performance of the first and the second device, the performance may comprise a measure of the key bit generation rate resulting from a QKD session between the first device and the second device. The inputs of the machine learning algorithm may comprise all of the factors (i)-(iii).

The machine learning algorithm may comprise a neural network. The method may comprise a preliminary step of training the algorithm using training data. The training step may comprise running the algorithm using measurements of factors (i)-00 as training inputs to the neural network and may comprise producing an estimate of a quantum key generation bit rate as an output of the neural network. The method may comprise running multiple training steps sequentially, each using different training data. As would be familiar to the skilled person, each of the multiple training steps may involve adjusting the neural network weights using backpropagation.

The method may comprise running the algorithm in respect of multiple pairs, each pair comprising the same satellite-based device and a different earth-based device. The method may further comprise ranking the earth-based communication devices in order of estimated key generation bit rate. The method may comprise taking the demand for quantum keys currently being experienced by a particular earth-based device into account when performing the ranking step.

In some embodiments the QKD system comprises one or more further pairs. The method may further comprise running the machine learning algorithm wherein the factors (i)-(iii) relate to the one or more further pairs. The one or more further pairs may comprise a different satellite-based communication device and the same earth-based communication device as that of first pair. The method may further comprise performing a comparison of the output of the algorithm in respect of the first pair with the output of algorithm in respect of the one or more further pairs. The method may further comprise using the outcome of this comparison to select one of the satellite-based devices to perform QKD with the earth-based device. In these embodiments, the method further comprises ranking the satellite-based communication devices in order of preference. This ranking step may take into account the estimated key generation bit rate between a ground-based device and a satellite-based device. This ranking step may also take into account the key demands of the ground-based device.

In some embodiments the second device may be a quantum receiver and may be located on the earth. In these embodiments, the first device receives the plurality of photons from a transmitter based on the ground and directs the plurality of photons to the second device.

In some embodiments the first device is the device located aboard the satellite. In these embodiments the second device may be located aboard a second satellite.

In some embodiments the machine-learning algorithm is a component of a larger algorithm, which may comprise deterministic functions.

The method can be performed by one or a combination of the first device and the second device.

According to a second aspect of the invention there is provided a QKD system, the QKD system being adapted to perform a QKD session that involves the propagation of a plurality of photons from a first device to a second device, one of the first and second devices being located aboard a satellite in orbit, the QKD system being adapted to run a machine learning algorithm, the inputs of the machine learning algorithm comprising one or more values of one or more of the following factors: (i) Atmospheric conditions in the vicinity of a path from the first device to the second device; (ii) The spatial separation of the first device and the second device; and (iii) A measure of the performance of the first and/or the second device, wherein the output of the algorithm is an estimate of the rate at which the pair generate bits of a quantum key.

A preferred embodiment of the invention will now be described in detail, for illustration only, and with reference to the appended drawings in which: Figure 1 is a schematic representation of a satellite capable of communicating with three earth stations each containing a QKD receiver; Figure 2 is a table showing typical training data that could be used in a method in accordance with the invention; Figure 3 is a flow chart showing the steps of a method in accordance with the invention.

Fig 1 shows a Low Earth Orbit (LEO) satellite 1. The satellite carries a quantum transmitter suitable for performing QKD (not shown). I will refer to the quantum transmitter as Alice 1. As the satellite progresses along its orbit, it comes within communicating range of one or more earth stations 2,3,4. By "earth station" it is meant an earth-based quantum communication device On this case a quantum receiver), located on earth 10. The earth stations shown in Fig. 1 are buildings 2,3,4. Each of the buildings contain a quantum receiver, which I will refer to as Bob. The satellite 1 performs a QKD session with the earth station 2 in order to distribute a plurality of quantum keys between the satellite 1 and the earth station 2. This QKD session uses the E91 QKD protocol. Specifically, a series of entangled photon pairs are produced aboard the satellite 1. One photon of each pair is directed into a detector aboard the satellite 1. The other photon of each pair is transmitted through the atmosphere to the earth station 2.

As is familiar to a person skilled in the art, each of Alice 1 and Bob 2 measure its respective incoming photons in one of a plurality of non-orthogonal basis states. This is known as the quantum transmission step. There then follows a key agreement step, in which Alice 1 and Bob 2 send each other, via a classical channel, an indication of which basis states each used to measure their photons. Alice 1 and Bob 2 then discard the measurements they made in respect of the photons for which they used different basis states. This leaves Alice 1 and Bob 2 with the same list of measurements (i.e. l's and O's) which constitutes a shared secret key.

This process is repeated multiple times so as to generate a plurality of such secret keys.

The number of keys that Alice 1 and Bob 2 successfully generate per unit time depends on various factors. These factors comprise: (i) weather conditions, such as atmospheric pressure and humidity; light pollution, such as light from buildings and vehicles; (iii) the separation of Alice from Bob -i.e. the straight-line distance between them); (iv) the performance of Alice and Bob -i.e. the rate at which Alice and Bob are able to accurately measure incoming photons.

While the QKD session described above is taking place, Alice 1 and Bob 2 each measure and record data values for the above factors. The rate at which Alice 1 and Bob 2 successfully generate quantum keys during the session is also recorded.

Alice will, at some point in the future, transmit the keys she now has as a result of the QKD session to a third party (not shown) so that that the third party can communicate in encrypted form with the earth station 2 using the keys. If Bob 2 transmits his keys to a fourth party (not shown), the third and fourth parties could use the keys for secure encrypted communication. In passing on the secure keys to other parties Alice and Bob have acted as trusted nodes.

Following termination of the QKD session between Alice 1 and Bob 2, Alice 1 initiates a QKD session with another Bob 3. Bob..3 is located in earth station 3. It is desired that this second Bob 3 also establish a plurality of quantum keys with Alice 1. Therefore, Alice 1 and the second Bob 3 perform a QKD session in the manner described above in relation to the first Bob 2. As in that session, Alice 1 and the second Bob 3 record data values for the above-mentioned factors (i)-(iv), along with the rate at which Alice 1 and the second Bobs successfully generate quantum keys during the session. As in the first session, the resulting quantum keys can then be distributed to other entities to facilitate secure quantum-encrypted communication.

Alice 1 repeats this process of performing QKD and recording data values for the above factors. The recorded data values are recorded in a central database, along with similar recorded values from other satellites. Over time, a large quantity of such data values in respect of factors (i)-(iv), along with the corresponding quantum key generation rates, in relation to multiple satellites, are accumulated in the database.

During much of a QKD satellite's orbit the on board Alice 1 is within range of multiple Bobs 2,3,4, each of which desires to perform QKD. (The satellite need not be directly overhead in order to perform QKD with a Bob). The present invention enables Alice 1, at any given time, to choose which Bob 2, 3, 4 will give the optimal QKD performance in view of the data values for factors (i)-(iv), prevailing at the time. The present embodiment of the invention achieves this using a machine-learning algorithm.

The purpose of the algorithm is to use the values of factors (i) to (iv) that are prevailing at a given time in order to estimate a value for the key generation rate obtainable during QKD between Alice 1 and a particular Bob 2,3,4. Alice 1 can, for example, use the algorithm to obtain quantum key generation rate estimates for each of the possible Bobs with which she could perform QKD, and perform QKD with the Bobs in order of the estimated rates, starting with the highest. Such an algorithm is achieved as follows.

The data values recorded in the database in respect of factors (i)-(iv) constitute training data for the machine learning algorithm. Typical examples of training data are shown in the table in Fig. 2. In particular, each of a given set of values obtained during one session, e.g. humidity, separation of Alice 1 from a given Bob 2,3,4, is used as an input to the algorithm. The algorithm uses randomly-chosen weighting values and produces an initial output. This is compared to the key generation rate that was measured by Alice 1 in respect of the QKD session in question. A process of back propagation is then performed which makes small adjustments to the weights of the neural network in order to more closely align the algorithm output with the key generation rate, in a manner which would be familiar to the skilled person. This process is repeated for all the training data. Thereafter, before Alice 1 begins a QKD session, she measures values for the factors (D-(iv) in respect of each of the Bobs 2,3,4 which are close enough to perform QKD with. The algorithm will produce an estimate for each Bob 2,3,4 ranking the Bobs 2,3,4 in order of their estimated key generation rate. This process is shown in the flow chart in Fig 3.

The optimisation process also takes into account the various key demands of the different available Bobs 2,3,4. It may be that one of the Bobs 2,3,4 has a very high key demand, but the algorithm estimates a relatively low key generation rate for it. The optimisation process applies a key demand-dependent weighting to that particular Bob, ensuring that its high key demand acts to move that Bob up Alice's priority rankings. Alice then performs QKD with the different Bobs in order of their position in the rankings.

Claims

Claims 1.A computer-implemented method of optimising a QKD system, the QKD system being adapted to perform a QKD session that involves the propagation of a plurality of photons from a first device to a second device, one of the first and second devices being located aboard a satellite in orbit, the method comprising running a machine learning algorithm, inputs of the machine learning algorithm comprising one or more values of one or more of the following factors: (i) atmospheric conditions in the vicinity of a path from the first device to the second device; (ii) a spatial separation of the first device and the second device; and (iii) a measure of the performance of the first and/or the second device, wherein the output of the algorithm is an estimate of the rate at which the first device and the second device generate bits of a quantum key.
2. A method as claimed in claim 1, wherein the first device is the device located aboard the satellite.
3. A method as claimed in claim 1 or claim 2, wherein the second device is located on the earth.
4. A method as claimed in any preceding claim, wherein the inputs of the machine learning algorithm comprise one or more values of factor (i), the atmospheric conditions in the vicinity of a path from one device of the pair to the other comprising air pressure, humidity, cloud coverage and pollution.
5. A method as claimed in any preceding claim, wherein the machine learning algorithm comprises a neural network.
6. A method as claimed in any preceding claim, wherein the method further comprises a preliminary training step of running the algorithm using measurements of factors (i)-(iii) as training inputs to the neural network and may comprise producing an estimate of a quantum key generation bit rate as an output of the neural network.
7. A method as claimed in claim 6, wherein the method further comprises running multiple training steps sequentially, each using different training data, each training step comprising adjusting neural network weights using backpropagation.
8. A method as claimed in any preceding claim, wherein the method comprises running the algorithm in respect of multiple pairs, each pair comprising the same satellite-based device and a different earth-based device and ranking the earth-based communication devices in order of estimated key generation bit rate
9. A method as claimed in claim 8, the method comprising taking a demand for quantum keys currently being experienced by a particular earth-based device into account when performing the ranking step.
10. A QKD system, the QKD system being adapted to perform a QKD session that involves the propagation of a plurality of photons from a first device to a second device, one of the first and second devices being located aboard a satellite in orbit, the QKD system being adapted to run a machine learning algorithm, the inputs of the machine learning algorithm comprising one or more values of one or more of the following factors: (iv) Atmospheric conditions in the vicinity of a path from the first device to the second device; (v) The spatial separation of the first device and the second device; and (vi) A measure of the performance of the first and/or the second device, wherein the output of the algorithm is an estimate of the rate at which the pair generate bits of a quantum key.
11. A computer program adapted to perform the method as claimed in claim 1.
12. A data carrier medium comprising the computer program as claimed in claim 11.