Fortifying AI: Tackling Adversarial Threats and Building Defenses

IEEE WiE East Tennessee is hosting a webinar Fortifying AI: Tackling Adversarial Threats and Building Defenses by Dr. Pravi Devineni is a Lead AI Scientist at Duke Energy. Abstract: AI systems now influence high-stakes decisions across sectors. Securing them requires a clear view of where vulnerabilities arise across the ML lifecycle. This session introduces common weaknesses from data collection and training through deployment and operations, and explains major adversarial threats—poisoning, evasion, inversion, and extraction, with real-world context. We’ll focus on practical detection and mitigation techniques and how these controls align with broader enterprise security practices. Attendees will leave with a simple framework and checklist to assess risk and prioritize safeguards. Learning objectives: – Map security risks across the AI lifecycle (data → training → validation → deployment → operations) – Explain key attack types—poisoning, evasion, inversion, extraction—and their impacts. – Apply defenses: data integrity and provenance checks, secure/robust training, evaluation and monitoring, incident response, red teaming, and governance controls. – Connect AI security work to enterprise security (threat modeling, identity, logging, SDLC). Bio: Dr. Pravallika (“Pravi”) Devineni is a Lead AI Scientist at Duke Energy. She designs secure-by-design, auditable AI systems for critical infrastructure, with a focus on grid reliability and nuclear operations. Her work spans adversarial ML Threat Modeling, evaluation and monitoring, and the practical application of NIST AI RMF and ISO/IEC 42001 to enterprise controls and evidence. Previously, she was a research scientist at Oak Ridge National Laboratory. She writes and speaks on AI risk and governance and mentors young scientists. She holds a Ph.D. in Computer Science from UC Riverside. Virtual: https://events.vtools.ieee.org/m/501800