A machine-learning model has been designed to predict emerging computer-network security threats with greater reported accuracy than support-vector-machine systems, according to research in the International Journal of Intelligent Information and Database Systems.
The approach described uses a radial basis function neural network (RBFNN). This is an AI, artificial intelligence, model designed to identify complex, non-linear relationships in data. The researchers trained the RBFNN on processed network-security information, including data that can be generated by intrusion-detection systems, firewalls and network traffic. The system ultimately makes network defence more proactive rather than relying on the predominantly reactive technology of firewalls and after-the-fact intrusion-detection systems. Instead of relying on predefined rules to recognise known threats, it can analyse incoming data and use its training to spot similar patterns and so act as an early-warning system.
In tests, the RBFNN correctly classified more than 95% of normal network activity and Heartbleed attacks and achieved 97% accuracy in identifying , denial of service (DoS) attacks. Its accuracy for brute-force attacks was somewhat lower, at up to 93%. However, it had an overall misclassification rate of less than 5%.
The researchers envisage a real-time service capable of automatically analysing data, detecting anomalies and generating alerts. Such systems could become increasingly important as attacks and advanced persistent threats become more varied and difficult to identify using fixed rules.
Liu, Y. (2026) ‘Prediction of computer network security situation based on machine learning’, Int. J. Intelligent Information and Database Systems, Vol. 18, No. 7, pp.1–18.
News media may use this press release as source material, in whole or in part, provided the content is not materially misrepresented. A link back to the original article is appreciated.
No comments:
Post a Comment