Advanced AI solution for enterprise malware detection

Overview _

Client:
ProLion GmbH
Cybersecurity, anti-malware solutions
Vienna, AT

Business case:
  • Detect and prevent ransomware attacks
  • Monitor the storage usage

Industry _

  • IT Services
  • Data centers
  • Storage
  • Cybersecurity

Services _

  • Custom software development
  • Product development

Project type _

  • Web
  • Distributed backend

Technology _

  • Java
  • NetApp Clustered Data ONTAP
  • Hazelcast
  • REST endpoints
  • AWS virtualisation
  • Machine learning

Description _

An application that runs on an enterprise’s datacentre storage systems (SAN) and does real time monitoring of the SAN I/O activity. It employs an advanced AI solution for anomaly detection that allows blocking malware from corrupting enterprise data.

Challenges _

Since malware can hit in many different forms and have a heavy impact on the final user, we have to:
_

  • Provide a powerful solution that protects against all threats (both known and new / unknown).
  • Ensure the best malware detection accuracy, while keeping false positives at minimum (or zero).
  • Deliver real-time detection and protection that spans across the whole SAN network.
  • Keep SAN performance unaffected.

Solutions _

We met client’s high expectations with a series of cross-technology solutions:
_

  • AI anomaly detection techniques that determine what is “normal” traffic and allow it to pass while “suspicious” traffic is blocked.
  • Model training and evaluation with extensive real data, collected from production SAN logs.
  • Processing and enhancement of collected data set, to obtain an even greater synthetic “real-like” dataset.
  • Setting up of simulated SAN environments; and release of malware to collect footprints.
  • Model parameters tweaking, to ensure highest precision and recall scores.
  • Implementation of distributed architecture, with sensors on each SAN node and dedicated processing nodes to run the detection model.
  • Development of a home-grown decision tree variant that is both accurate and lightweight enough for the use case.
  • Hyperparameter tuning to minimize the model while maintaining the accuracy.

29 years in business | 2700 software projects | 760 clients | 24 countries

We turn ideas into software. What is yours?

Get in touch

10 + 11 =