Continuous Learning

Enhancing an Intelligence Analysis Firm's Accuracy in Identifying Persons and Posts of Relevance to Cybersecurity

Big data analytics in cyber defense refers to the analysis of vast amounts of data to predict cyber attacks. An Intelligence Analysis firm used the Engine to gain predictive insights to power their cyber-surveillance and intelligence platform.

Social media platforms are a growing major source of news and information. Within that constant stream of cyber data are business insights and intelligence that could help identify persons of interest to investigate and prevent malicious activity.

PI.EXCHANGE partnered with an Intelligence Analysis firm (the Client) to power their cyber-surveillance and intelligence platform with predictive insights. Delivering a solution required a means to automatically analyze large volumes of cyber-surveillance data, as well as automatically predict, identify and triage potential "matters & persons of interest" in relation to potential malicious intent over time. 


The objective: To identify a solution that could respond to the dynamic and changeable nature of cyber data over time, by enabling continuous learning of deployed models. This meant that the accuracy of predictive output from deployed models iteratively improved, and remained relevant as new labelled data was provided to The Engine. 


The Client utilizes a platform for their internal team to perform keyword searches as part of a service offering social media cyber intelligence to their clients. These inputs return a huge amount of cyber data. This requires a significant manual investigative task for the client’s team, which is slow and prone open to manual handling errors. The Engine is able to predict a relevance score against the keyword,  proffering a subset of highly relevant data back to the internal users. 


Using PI.EXCHANGE’s AI & Analytics Engine, the Client was able to carry out big data analytics in cyber defense. That is, the Client could quickly model vast amounts of historical cyber data, then stream in new data to identify posts and account aliases of interest, with the sole objective of predicting and prevent cyber attacks.

Historical social media data was labelled, then prepared within the AI & Analytics Engine for the purpose of developing a predictive model. The Engine recommended the optimal model given the prepared dataset and target column. The Engine can uniquely suggest the best performing model prior to training. This model was selected, then trained and tested against the labelled training data. 

With a single click, the model was deployed on a protected VPC in accordance with the high-security requirements. From this point, the model, deployed to the endpoint could be called on via the client's GUI for predictive output against input cyber data. In effect, providing a highly relevant subset of "matters & persons of interest" in relation to potential malicious intent for their team to provide their clients. 

To make sure the deployed model remained relevant to the incoming stream of data, the models had “continuous learning” switched on. A feature that allows data augmentation, where new labelled data could be provided so that the model remains relevant and can improve its accuracy. 

“By leveraging the smart and highly scalable automation and pre-tuned technology available within their AI & Analytics Engine, we were able to significantly accelerate insight extraction whilst maintaining the highest standards of security & privacy in an extremely cost effective setup."  

Chief Operating Officer


We delivered a fully integrated prediction API with continuous learning capability in 10  days, enabling the following:

  • Reduced total development cost;
  • Lowered ongoing maintenance cost (minimal human interaction required); 
  • Accelerated insight extraction & analysis; and
  • Continuous improvement and sustained predictive power and relevancy of deployed models.



The solution provided an automated predictive engine that enhanced the accuracy and reduced the manual investigative hours spent analyzing data for clients. 

By utilizing the continuous learning feature, the client provides new labelled data back to the Engine and, over time, benefits from incremental improvement in the accuracy of the model and a continuously relevant model given the changing data. Continuous learning can simply be switched on at the time of model deployment.

Continuous learning on the AI & Analytics Engine 

The value of PI.EXCHANGE’s AI & Analytics Engine extends beyond the immediate application. The Engine as a tool, empowers internal teams to support a sustainable  AI/ML capability, providing an additional competitive edge in the competitive and rapidly evolving cybersecurity industry.

Wrap Up

Using the AI & Analytics Engine, the Client is able to build a repeatable collaborative ML development workflow for their team. In an industry with time-sensitive data and rapid changes, the ability to quickly develop models specific to their use cases and iteratively improve them with new labelled data is imperative. Read on for more AI/ML industry solutions

Book a demo with us to find out how you can derive meaningful insights from your data!

Book a demo 

Similar posts

Subscribe and get the latest news, articles, and thought pieces direct to your inbox.