By leveraging machine learning techniques in its internal service, GReAT uncovered thousands of new advanced threats targeting government, finance, enterprise, and telecommunications sectors.
These findings were achieved by analyzing global cyberthreat data from the Kaspersky Security Network (KSN).
The machine learning models employed in Kaspersky’s solutions use techniques such as Random Forest and term frequency–inverse document frequency (TF-IDF) to process vast amounts of data, enabling faster and more accurate detection of subtle threats.
This combination of ML methods allows for the identification of indicators of compromise (IoCs) that traditional detection systems might overlook, leading to more precise anomaly detection and a significant improvement in overall threat detection capabilities.
Kaspersky’s ongoing use of machine learning has allowed its systems to process millions of data points daily, providing real-time insights into emerging threats.
This has resulted in a 25% increase in threat detections for the first half of 2024, significantly enhancing the ability to reduce response times and mitigate cyber risks.
“The results have exceeded our expectations,” said Amin Hasbini, Head of the META Research Center at Kaspersky’s GReAT.
“These technologies improve detection accuracy and foster proactive defense strategies, helping organizations stay ahead of evolving cyber threats.
The future of cybersecurity lies in ethically harnessing these tools to ensure a safer digital environment for all.”
Kaspersky’s machine learning models are regularly refined and updated with new data to maintain their effectiveness in a constantly changing cyberthreat landscape.
As new attack vectors emerge, these models are monitored and tuned to provide timely insights, strengthening defenses and enhancing organizational resilience.
The research results will be discussed at GITEX 2024, where Kaspersky will participate in a panel on the impact of AI on cybersecurity.
To learn more about machine learning in threat hunting, visit Securelist.com.