The recent advances in Artificial Intelligence (AI) and machine learning (ML) algorithms drive the proliferation of data-driven intelligent applications in many areas including advanced command and control (AC2) missions, decision support systems, recommender systems, computer vision, autonomous vehicles, as well as intrusion detection and prevention systems. Many AI/ML algorithms have been integrated with decision support applications to improve analysis performance and prediction. Due to the widespread usage of AI/ML in critical decision processes (e.g., DoD AC2 mission critical applications), there is an exponential growth in cyberattacks that target the AI/ML algorithms and consequently influence their decision process in favoring the attackers, said Dr. Salim Hariri of LOCH Technologies.
Revolutionizing Cyber Protection of AI/ML Algorithms with Resilient Machine Learning Systems (rMLS)
Today, Dr. Salim Hariri and his research team are at the forefront of developing cutting-edge solutions that make AI/ML systems tolerate any type of attack against them. AVIRTEK, a subsidiary of LOCH Technologies awarded $1.8M from the Air Force Research Laboratory (AFRL) for developing Resilient Machine Learning Systems (rMLS) that continue to operate normally despite being attacked by adversarial AI/ML attacks. In this project, according to Dr. Hariri, our focus is on securing the AI/ML system as opposed to using the AI/ML system to build security measures for a wide range of applications in DoD and commercial markets. The researchers at AVIRTEK and LOCH will develop a resilient ML system (rMLS) that ensures resilient AI/ML operations from any type of cyber attack. Traditional defense mechanisms fall short of protecting AI/ML systems from cyber attacks, said Dr. Hariri. “Our rMLS represents a revolutionary approach to protect AI/ML systems by utilizing Moving Target Defense (MTD), where the rMLS dynamically changes the ML algorithms used. By randomly changing the ML models, we make it extremely difficult for the attackers to know our internal AI/ML algorithms, and consequently ensuring resilience to adversarial AI/ML attacks.”
Looking Ahead: rMLS Technology and Broader Impacts
The rMLS technology will have a broad impact because currently, most commercial AI/ML platforms have not addressed the security issue of their AI/ML algorithms, said Dr. Hariri. Consequently, the rMLS technology will have huge marketing opportunities in two markets:
“As AI continues to advance, Large Language Models (LLMs) such as GPT-4 and BERT pose an additional layer of threat. LLM-based attacks can generate sophisticated phishing schemes, alter the interpretation of data, or compromise ML decision-making. The rML system’s MTD-based approach is uniquely suited to counter these emerging threats, making it a critical tool in defending against LLM adversarial attacks. By continuously evolving to meet the challenges posed by next-generation AI threats, rML systems offer the cybersecurity industry a vital edge,” said Garry Drummond, Founder and CEO of LOCH Technologies.
For more information on this research and future developments, please visit www.loch.io/acs