This track addresses the growing need for secure and efficient AI processing on embedded multicore system-on-chip (SoC) platforms. As AI workloads become pervasive in edge and IoT devices, ensuring data privacy while maintaining high performance is a critical challenge. This track invites research and development efforts that integrate privacy-preserving techniques with hardware acceleration strategies tailored for embedded multicore environments. This track fosters interdisciplinary collaboration across AI, embedded systems, security, and hardware design communities to advance trustworthy and scalable AI deployment at the edge.
Topics of Interest Include:
- Privacy-preserving AI methods:
- Federated learning, differential privacy, and homomorphic encryption
- Secure multi-party computation and trusted execution environments
- Hardware acceleration for AI:
- Custom accelerators for privacy-aware inference and training
- Optimization of AI workloads on multicore and manycore SoCs
- Energy-efficient and real-time AI processing
- System-level integration:
- Co-design of hardware/software for secure AI pipelines
- Memory and communication architectures for privacy-aware AI
- Secure boot and runtime protection for AI-enabled SoCs
- Evaluation and benchmarking:
- Performance metrics for privacy-preserving AI
- Trade-off analysis between privacy, accuracy, and efficiency
- Comparative studies of secure AI frameworks on embedded platforms
