Cognitive Cybersecurity Intelligence (CCSI) Group - overview

The Cognitive Cybersecurity Intelligence (CCSI) group (formerly Global Security Analysis Lab or GSAL) researches on methodologies and technologies to help organizations detecting, understanding, and deflecting advanced cyber security threats and attacks on their network and in the cloud. It explores challenging research problems posed by building and combining AI and cognitive methods (e.g., contextual and behavioral analysis, machine learning, reasoning), scalable big data security analytics (e.g., graph mining, deep correlation and provenance analysis), and next-generation defense mechanisms (e.g., transparent malware analysis, active defense and cyber deception layers) to gain deep intelligence and insights about cyber security threats and attacks as well as threat actors. 


Current focus areas and projects

  • AI-powered and cognitive security offense analytics, cyber threat hunting, and threat intelligence consolidation
  • Cross-stack cyber deception and active defense techniques
  • Cyber security analytics, event correlation, and provenance tracking on the network and device-level
  • Next-generation malware analysis
  • Design of high-speed and scalable data collection platforms for real-time and historical security analytics
  • Security data visualization and penetration testing


Recent Highlights


Recent Impact on IBM Products and Offerings

The CCSI had direct impact with core contributions to several new security products and solutions IBM launched in the last few years:


On-going Projects and Efforts

Cognitive Security Analytics and Threat Intelligence

We are researching and developing techniques and methodologies to apply cognitive analytics and IBM Watson technologies to challenging security problems. Our research is the foundation of the release of Watson for Cyber Security by IBM Security in 2016 and IBM QRadar Advisor with Watson in 2017.

AI-powered Big Data Cyber Security Analytics

We explore and develop novel security analytic methods that deliver sustainable cyber security defenses against emerging advanced and persistent threats (e.g., deploying data mining and machine learning techniques to detect benign, suspicious, and malicious behaviors across several heterogeneous data channels).

Active Cyber Deception and Defense

We research on methodologies, techniques, and technologies to build cyber deceptive systems on multiple layers of an organizations' IT stack with the goal of detecting and deflecting adversarial activities and thereby make adversaries reveal inadvertently their presence, capabilities, and intentions.

Feature Collection and Correlation Engine

Design, architecture, and implementation of a novel analysis engine, called FCCE, which finds correlations across a diverse set of data types spanning over large time periods with very small latency and with minimal access to raw data. Our engine scales well to collecting, extracting, and querying features from geographically distributed large data sets at close-to-real-time or from historical data sets.

Malware Analysis, Ethical Hacking, and Penetration Testing

Next-generation malware analysis technologies, Security Threat and Vulnerability Analysis, Ethical Hacking, Network Forensics, etc.



Jiyong Jang

DeepLocker: How AI Can Power a Stealthy New Breed of Malware

Exploring the Security Knowledge Graph

Security Knowledge Graph

Identify and Understand threats with Watson for Cyber Security