Session
Hidden in Plain Sight: Designing Systems for Private Intelligence
DescriptionThis session explores the state of confidential AI systems from fully homomorphic encrypted inference to secure orchestration pipelines and concludes by examining emerging hardware and system-level attacks that challenge these protections.
Event Type
Research Manuscript
TimeMonday, July 2710:30am - 12:30pm PDT
LocationMtg Room 203C
Security
SEC1. AI/ML Security/Privacy
Presentations
| 10:30am - 10:43am PDT | Securerouter: Encrypted Routing for Efficient Secure Inference | |
| 10:43am - 10:56am PDT | TT-SEAL: TTD-Aware Selective Encryption for Adversarially-Robust and Low-Latency Edge AI | |
| 10:56am - 11:10am PDT | FBS: Accelerating CNN Inference over RNS-CKKS with Fewer Bootstrapping Sparsity | |
| 11:10am - 11:23am PDT | Runtime-ADAR: Runtime Anomaly Detection and Attack Recovery in Encrypted DNNs | |
| 11:23am - 11:36am PDT | Holocode: Hybrid Optical-Electronic Edge Encoding for Privacy-Preserving Cloud Training | |
| 11:36am - 11:50am PDT | PLONK-Hammer:breaking Input Privacy of PLONK Proving Systems via Rowhammer | |
| 11:50am - 12:03pm PDT | MACH: Memory-Aware Configuration Generation for Homomorphically Encrypted Neural Networks | |
| 12:03pm - 12:16pm PDT | Icarus's Wings: Disabling MoE Offloading Acceleration via a Universal Hidden Prefix Attack | |
| 12:16pm - 12:30pm PDT | Multi‑GPU Tensor‑Level Fusion and Adaptive Memory Management for Private Inference |