GLAMP: Generative Learning for Adversarially-Robust Malware Prediction

NU 2025-002

INVENTORS

• Venkatramanan Subrahmanian* (McCormick School of Engineering, Computer Science)
• Cristian Molinaro
• Lirika Sola Or Sholla
• Saurabh Kumar

SHORT DESCRIPTION

For cybersecurity firms, GLAMP is a framework that proactively generates novel malware variants to train detectors and improve defense against future attacks.

BACKGROUND

Traditional antivirus systems struggle with evolving threats and limited malware samples. Current solutions react only after malware is identified. This creates a high cost and lag in defense, leaving a significant gap in proactive cybersecurity strategies.

ABSTRACT

GLAMP addresses the challenge of adaptive malware by generating variants of known malware. The framework formalizes the malware generation problem and integrates novel variant generation algorithms with an adversarial training model. Experiments show that GLAMP successfully evades 11 white box classifiers and 4 commercial detectors. The system enhances malware prediction by exposing classifiers to both historical and generated malware samples.

MARKET OPPORTUNITY

The global market for Endpoint Detection and Response (EDR) is a direct response to the failures of traditional antivirus and was valued at approximately $4.1 billion in 2024. It is projected to reach $11.85 billion by 2029, growing at a rapid compound annual growth rate (CAGR) of 23.65%. This growth is fueled by the escalating volume and sophistication of zero-day exploits, fileless malware, and ransomware that easily bypass legacy signature-based systems. The primary market consists of enterprises, mid-market businesses, and government agencies seeking to close the critical security gap left by reactive solutions. (Source: Mordor Intelligence: "Endpoint Detection and Response (EDR) Market Size & Share Analysis - Growth Trends & Forecasts (2024 - 2029)").

DEVELOPMENT STAGE

TRL-4 - Prototype Validated in Lab: A laboratory-scale prototype has demonstrated key functions, including novel malware generation and successful evasion of multiple detection systems.

APPLICATIONS

• Proactive malware variant generation: Trains detectors with unseen samples to boost resilience.
• Enhanced threat anticipation: Empowers cybersecurity firms to predict and counter future attacks.

ADVANTAGES

• Improves detection robustness: Exposes classifiers to a wider range of malware variants.
• Enables proactive defense: Anticipates and counters emerging threats before they occur.

PUBLICATIONS

• Venkatramanan Subrahmanian et al., GLAMP: Generative Learning for Adversarially-Robust Malware Prediction IEEE, 10 July 2025

IP STATUS

US Patent Pending