Uploaded August 2026 | Updated September 2026, 11 hours ago
Sneha Rangari (Visa, US)
GPUs power the AI revolution and silently harbor the next wave of cyber threats. From data centers to cloud AI workloads, attackers are exploiting GPU memory, kernels, and interconnects in ways traditional security tools cannot detect.
As AI reshapes industries, Graphics Processing Units (GPUs) have become indispensable engines, accelerating everything from neural network training to high‑performance inference. Yet this reliance has created a new, largely invisible attack surface. Adversaries are leveraging this gap to deploy VRAM‑resident malware (Jellyfish), perform GPU snooping, launch side‑channel attacks, hijack GPUs for cryptocurrency mining, and even steal proprietary AI models — all while evading conventional defenses.
This session draws on primary research, case studies, and hands‑on demonstrations to illustrate how GPU threats compromise AI confidentiality, integrity, and reliability. Attendees will gain actionable techniques for GPU telemetry monitoring, VRAM forensics, and integrating GPU visibility into SOC workflows.
The talk also empowers participants to take ownership of GPU security in their own environments, exploring research, defense strategies, and leadership opportunities in hardware‑aware AI protection. By the end, participants will understand why securing GPUs is critical to securing the future of AI — and how to illuminate the dark silicon powering intelligent systems worldwide.
---
Sneha Rangari is a Senior Cybersecurity Engineer at Visa with over eight years of experience specializing in security architecture, cloud platforms, and enterprise risk management. A CISSP and GMLE certified professional, she focuses on securing emerging technologies through AI/ML integration, including developing advanced Deepfake detection and threat modeling frameworks. Beyond her technical role, Sneha is an active mentor with Women in CyberSecurity (WiCyS) and a frequent speaker at industry conferences like RSA and BSides. She is dedicated to fostering an inclusive security community and aligning technical defense strategies with global business priorities.
Sneha Rangari (Visa, US)
GPUs power the AI revolution and silently harbor the next wave of cyber threats. From data centers to cloud AI workloads, attackers are exploiting GPU memory, kernels, and interconnects in ways traditional security tools cannot detect.
As AI reshapes industries, Graphics Processing Units (GPUs) have become indispensable engines, accelerating everything from neural network training to high‑performance inference. Yet this reliance has created a new, largely invisible attack surface. Adversaries are leveraging this gap to deploy VRAM‑resident malware (Jellyfish), perform GPU snooping, launch side‑channel attacks, hijack GPUs for cryptocurrency mining, and even steal proprietary AI models — all while evading conventional defenses.
This session draws on primary research, case studies, and hands‑on demonstrations to illustrate how GPU threats compromise AI confidentiality, integrity, and reliability. Attendees will gain actionable techniques for GPU telemetry monitoring, VRAM forensics, and integrating GPU visibility into SOC workflows.
The talk also empowers participants to take ownership of GPU security in their own environments, exploring research, defense strategies, and leadership opportunities in hardware‑aware AI protection. By the end, participants will understand why securing GPUs is critical to securing the future of AI — and how to illuminate the dark silicon powering intelligent systems worldwide.
---
Sneha Rangari is a Senior Cybersecurity Engineer at Visa with over eight years of experience specializing in security architecture, cloud platforms, and enterprise risk management. A CISSP and GMLE certified professional, she focuses on securing emerging technologies through AI/ML integration, including developing advanced Deepfake detection and threat modeling frameworks. Beyond her technical role, Sneha is an active mentor with Women in CyberSecurity (WiCyS) and a frequent speaker at industry conferences like RSA and BSides. She is dedicated to fostering an inclusive security community and aligning technical defense strategies with global business priorities.
