Invention Description
Integrating hardware into safety- and mission-critical systems has sparked strong demand for combined edge and cloud solutions. This ecosystem helps extend product lifespan while maintaining service quality, driven largely by the Silicon Lifecycle Management (SLM) framework. Silicon Lifecycle Management (SLM) ensures hardware longevity and reliability using a coordinated framework of on-chip sensors, telemetry, and machine learning. While it optimizes performance and minimizes downtime, SLM integration faces major challenges from high overhead, resource constraints, and a heavy reliance on labeled datasets for analytics.
Researchers at Arizona State University have developed OCTANE, a statistical method which leverages on-chip telemetry through performance counters and sensors combined with an unsupervised, workload-independent anomaly detection mechanism called OCTANE-edge, which is optimized for compute and memory efficiency using fixed-point arithmetic. Extending detection capabilities, OCTANE-cloud provides an end-to-end anomaly-informed diagnosis model using compact telemetry signatures, enabling accurate security and safety incident diagnosis. Validated with the PAMPAR benchmark suite, OCTANE delivers superior detection accuracy and diagnosis performance with minimal area and power overhead.
OCTANE is an innovative on-chip anomaly detection and diagnosis system that enhances silicon lifecycle management by integrating efficient hardware/software co-design for real-time monitoring and analysis.
Potential Applications
- Silicon lifecycle management for semiconductor manufacturers
- Real-time anomaly detection in integrated circuit design and testing
- Security monitoring and threat detection on chip hardware
- Safety-critical system reliability enhancement in automotive and aerospace industries
- Embedded system health monitoring and predictive maintenance solutions
Benefits and Advantages
- Real-time on-chip anomaly detection with low power and area overhead
- Unsupervised, workload-independent anomaly detection with OCTANE-edge
- End-to-end diagnostic capability via cloud-based anomaly telemetry signatures
- Superior performance validated against methods like E-SCOUT, iForest, and autoencoders
- Efficient on-chip anomaly detection using hardware/software co-design
- Compute- and memory-efficient implementation using fixed-point arithmetic
For more information about this opportunity, please see