* Identify vulnerabilities across AI models, autonomous systems, and cyber-physical environments
* Analyze adversarial ML threats including evasion, poisoning, and model exploitation
* Apply threat modeling techniques tailored to interconnected and real-time systems
* Design security controls that address both digital integrity and physical safety
* Implement monitoring, anomaly detection, and autonomous response strategies
* Reduce risk by understanding how cyber attacks translate into physical consequences
* Strengthen protection of critical infrastructure including transportation, utilities, and smart cities
* Mitigate supply chain risks across hardware, software, and AI components
* Support compliance with emerging regulations around cyber-physical and AI systems
The CA306 program explores the security of intelligent and cyber-physical systems, where software decisions directly impact the real world. It spans everything from AI model vulnerabilities to sensor-level attacks and large-scale interconnected infrastructure risks.
Understand how AI, autonomous systems, and cyber-physical systems merge digital and physical risk.
Examine how system architecture and trust relationships influence real-world security outcomes.
Shift into how intelligent models behave under adversarial conditions.
Analyze how small input changes can break systems that appear highly accurate.
Move into systems that interact with the physical world through movement and perception.
Evaluate how sensors, control logic, and communication channels become attack vectors.
GPS Spoofing Scenario Analysis
Sensor, Decision & Action chain Analysis
Expand into large-scale intelligent environments and interconnected systems.
Understand how dependencies and external components introduce systemic risk.
Conclude with defensive strategies for detecting and responding to system anomalies.
Focus on maintaining reliability even when systems are partially compromised.
Looking for Group Purchase Options? See below
Intelligent systems combine software, sensors, autonomous decision-making, and physical operations into a single operational chain. This means attackers can influence not only data, but also real-world actions such as movement, routing, industrial control, and autonomous behavior, turning cyber compromise into physical consequence.
Machine learning systems rely on patterns and statistical correlations rather than true understanding. Because of this, small manipulations in data or inputs can mislead models and cause incorrect decisions, even when the system appears to operate normally.
Adversarial attacks manipulate AI inputs in subtle ways to force incorrect predictions or classifications. These attacks can target images, text, sensor data, or sequential systems while remaining difficult for humans or traditional validation systems to detect.
Autonomous systems operate at a speed and scale where humans cannot constantly verify every action. Because of this, systems must rely on strong identity, hardware-backed trust, attestation, and continuous validation to ensure devices and services are authentic and operating in a trustworthy state.
Resilient systems are designed with the assumption that software and decision logic may eventually fail. Instead of relying entirely on software trust, resilient architectures use safeguards such as semantic integrity checks, hardware interlocks, safety systems, and independent emergency controls to prevent digital compromise from becoming uncontrolled physical impact.
