QuickHeal

We helped QuickHeal to build a Machine Learning powered Auto-Malware detector that scans tens of thousands of files/hour to identify and classify threats automatically.

It Saved them hundreds of man-hours per day with improved accuracy of malware detection by more than 70%.

The key challenges were -

• To gain competitive advantage by accurate prediction of malicious / non-malicious files using Machine learning
• Model size must be small
• Desired False Positive Rate (FPR) < 0.5 %
• Training on massive data (Tens of millions of data points)

CoreView team worked in an agile manner to deliver a state-of-the-art system.

• Advanced ML classification algorithm to predict malicious files
• Detailed feature engineering with a variety of file data
• Classifiers models like Gradient Boosting, Decision Tree, etc …
• Automated pipeline for data ingestion and training
• APIs for consumption, training, feedback of model
• Leaner model and very fast prediction, minimal compute
• Deployed with PySpark on AWS