IGLOO SECURITY, INC. builds 120 million AI datasets that reproduce the latest breaches KISA ‘Cyber Security AI Dataset (Infringement Incident Area)’ Construction Completion of Infringement Case Reproduction Field

- IGLOO Security completes building a dataset that reproduces the latest breaches based on AI capabilities accumulated over many years

- Utilization of AI datasets open to the private sector is expected to speed up AI-based intrusion response intelligence 

- IGLOO Security plans to strengthen its data business this year based on data-centric AI capabilities

[January 26, 2022] As the Artificial Intelligence (AI) dataset that reproduces the latest breaches is opened to the private sector, it is expected that more organizations will be able to enjoy the benefits of AI and strengthen their security systems. IGLOO Security, a leading artificial intelligence (AI)-based cybersecurity company (CEO Lee Deuk-Choon, www.igloosec.com/en) will build a ‘Cyber Security AI Dataset (infringement incident area)’ of the Korea Internet & Security Agency (KISA) in 2021. It announced that it has completed the construction of 120 million datasets that reproduced major breaches by participating in the project. 

As part of the K-Cyber Prevention Promotion Strategy, the Korea Internet & Security Agency (KISA) is building a ‘Cyber Security AI Dataset Construction Project’ to build more than 800 million AI datasets that can be applied to AI infringement response in the field of malicious code and intrusion incidents. The goal of this project was to prepare a basis for Korean security organizations to preemptively respond to new and variant security threats based on the dataset of breach incidents open to the private sector. IGLOO Security, which has been working hard to develop AI algorithms and AI learning data since 2015, has been in charge of reproducing the latest breaches in this project. 

IGLOO Security took on the role of building and verifying AI datasets that reproduced major global breaches based on years of experience in building and operating AI systems and large-scale threat response. IGLOO Security collects raw data generated from 6 heterogeneous security devices through the execution of 15 carefully selected breach scenarios, extracts and selects attack characteristics from the raw data, labels them, and then uses this labeling data as data for AI learning. After that, the AI dataset was applied to the site to which the AI security model was applied and verification was completed. 

IGLOO Security has also developed a ‘Playbook’ for 15 breaches so that Korean security organizations can systematically and quickly respond to various breaches such as attacks using web application vulnerabilities and ransomware infection. IGLOO Security analyzes the attack stage according to MITRE ATT&CK Framework and through the process of executing the penetration test from the attacker’s point of view, they prepared a ‘Playbook’ that manualized the optimal standard procedures and countermeasures for each type of intrusion incident. 

The dataset constructed this time will be opened to the public through the Cyber Security Big Data Center of the Korea Internet & Security Agency. By using this dataset, Korean security organizations are expected to dramatically reduce the time required for data collection, processing, and verification, and speed up AI-based intrusion response intelligence. In order for more Korean security organizations to enjoy the benefits of AI, IGLOO Security plans to focus on data voucher support project participation and data labeling tool development along with building a dataset in the field of intrusion incidents. 

Lee Deuk-Choon, CEO of IGLOO Security, said, “The true value of AI can only be realized when there is high-quality learning data that AI algorithms can learn well along with AI algorithms, in this regard, IGLOO Security has been accumulating AI learning data that can improve the accuracy and reliability of AI algorithms along with the development of AI algorithms specialized for security for many years. Based on such data-centric AI capabilities, this year along with the existing security business, we will accelerate the strengthening of the data business to provide high-quality learning data.”