Ayesa – Anomaly detection in cybersecurity attacks in industrial environments using quantum algorithms.
Ayesa
Sector: ICT
Business Case
Detection of anomalous behaviour in industrial safety systems as critical systems. How can we reduce training times and volumes of current unsupervised systems? How can we reduce the detection latency of possible attacks by improving the sensitivity of the models to unknown attacks?
Objectives
Use quantum parallelism (the possibility of applying an algorithm to a multitude of states in parallel), to obtain anomalous patterns that may be indicative of unknown attacks, with greater detail (specificity), and a lower latency time than classical anomaly detection algorithms.
Use case
In industrial cybersecurity contexts, and with the presence of a large amount of input data, some of which comes from unreliable or unknown sources, it is important to be able to detect outliers in the data at a very fast response pace. This is especially important when few outlier samples may be available to develop an effective ML model. These outliers may be indicative of some unexpected phenomena that arise in a system where they had never been identified, such as a faulty system or a malicious intruder. The quantum advantage of the application of unknown pattern detection algorithms over classical algorithms is mainly based on the possibility of quantum circuits to analyse all possible states of a system in polynomial and non-exponential times.
Infrastructure
Cloud
Technology
Quantum Automatic or deep learning
Data
Private datasets
Resources
Quantum Computational Scientist – Data Engineers – Cybersecurity Experts
Difficulties and learning
Development of integrated quantum algorithms, quantum PCA, quantum Encoders-Decoders. Scaling of the solution on current NISQ computers.
KPIs (business impact and metrics of the model)
It has been demonstrated that, through quantum technology and the algorithm used, it is possible to detect anomalies in classical data in the service, transport and industrial protocol layers in a more specific way (the quantum algorithm is capable of analysing more specific and particular types of anomalies than its classical counterparts) and with a lighter training than with traditional classical computing.
Funding
Q4Real Project: This project received funding from the competitive Hazitek research and innovation programme of the Basque Government.
Collaborators, Partners
ITS, Tecnalia.
