Page 56 - Kaleidoscope Academic Conference Proceedings 2021

3.2.4 QKDN - Applications of machine learning 3.2.5 QKDN - Resilience framework

ITU-T published Supplement 70 [19] to present the For resilience in a QKDN, Y.QKDN-rsfr specifies the
applications of ML in QKDNs as follows: framework of resilience in a QKDN including the conceptual
・ Applications of ML in the quantum layer of a QKDN: models of QKDN protection and recovery scenarios. It also
ML-based quantum channel performance prediction provides typical use cases of resilience and related
(QL01), QKD system parameter optimization (QL02), requirements of resilience schemes supported by the
and remaining use life (RUL) prediction of components quantum layer, the key management layer, and QKDN
in a QKD system (QL03). control and management layers, respectively. Y.QKDN-rsfr
considers the following typical scenarios of resilience in a
・ Applications of ML in the key management layer of a QKDN: 1) resilience in a QKDN supported by 1+1
QKDN: ML-based key formatting (KM01), key storage protection, 2) resilience in a QKDN supported by 1:1 or 1:n
management (KM02), and suspicious behavior recovery, and 3) resilience in a QKDN supported by re-
detection in the key management layer (KM03). routing.
・ Applications of ML in the control and management
layers of a QKDN: ML-based data collection and data 3.2.6 QoS aspects in QKDN
preprocessing (CML01), routing (CML02), and QKDN
fault prediction (CML03). There are three work items for QoS aspects in QKDN as
follows:
The ML pipeline subsystem in a QKDN is shown in Figure ・ Y.QKDN-QoS-pa: It covers the descriptions of QoS
8. The ML functional elements in the ML pipeline subsystem and network performance in a QKDN, classification of
include a Collector (C), a Preprocessor (PP), a Model (M), a performance concerns for which parameters may be
Policy (P) and a Distributor (D). The ML functions enable us needed, QoS parameters of a QKDN and network
to collect input data from the Source of data (SRC) through performance supporting factors.
data handling interfaces. The SRC can be in multiple layers
of QKDNs (see Figure 3). The target of the ML output (SINK) ・ Y.QKDN-QoS-fa: It gives an overview of QoS
can be elements in the quantum layer, the key management assurance for a QKDN, a functional architecture of QoS
layer and QKDN control and management layers. More assurance for a QKDN, reference points of functional
details related to the ML pipeline subsystems can be found architecture and procedures of QoS assurance for a
in [20]. QKDN.
ML pipeline subsystem ・ Y.QKDN-QoS-ml-req: It first provides an overview of
SRC ML Functions SINK requirements of ML-based QoS assurance for QKDN.
Service layer QKDN Control and It also describes a functional model of ML-based QoS
QKDN Control layer C PP Management layers assurance followed by associated high level and
QKDN Management layer functional requirements of ML-based QoS assurance.
Key Management layer M P D Key Management layer
Quantum layer Quantum layer
3.2.7 Standardization roadmap on QKDN
Data Handling interfaces Data Handling interfaces
QKDN Y.supp.QKDN-roadmap provides the standardization
C: collector; PP: preprocessor; M: model; P: policy; D: Distributor; SRC: source of data; SINK: target of ML output roadmap on QKDNs. It describes the landscape with related

technical areas of trust technologies from an ITU-T
Figure 8 – ML pipeline subsystem in a QKDN perspective and list of related standards and publications
developed in other SDOs.
For the functional requirements and architectures for an ML-
enabled QKDN, Y.QKDN-ml-fra specifies the roles of ML 4. PRE-STANDARDIZATION ACTIVITIES IN
in a QKDN. In particular, Y.QKDN-ml-fra includes ITU-T FG-QIT4N
functional requirements and a functional architecture model
of the ML-enabled QKDN. To specify the functional A Quantum Information Network (QIN or Quantum Internet)
architecture to enable ML in a QKDN, it applies the high- is expected to connect quantum information processing
level architecture specified in [20] to fulfill the requirements nodes, including QKD nodes, quantum computers and
for the ML-enabled QKDN. The functional architecture for quantum sensors, via quantum communication technologies
the ML-enabled QKDN includes three subsystems: QKDN- such as quantum teleportation and quantum repeating, to
related ML pipeline; QKDN-related ML sandbox; and realize quantum information transmission and networking.
QKDN-related Machine Learning Management Subsystem QIN has the potential to provide a series of new applications,
(MLMS). Further considerations on locations of ML-related such as distributed quantum computing and quantum sensor
functions will be discussed with detailed procedures. networks with the following technologies:
・ Quantum computing: a new computation model that
follows the laws of quantum mechanics to control
quantum information units.

– lii –