MACHINE LEARNING-DRIVEN DERIVATION OF ABSTRACT BEHAVIOURAL PATTERNS FROM RUNTIME ACTIVITY LOGS IN SYSTEM-ON-CHIP SYSTEMS
Bibliographic record
Abstract
System-On-Chip Systems platforms generate immense volumes of runtime activity logs that hold rich information about system behaviour, reliability, and fault conditions. Traditional log analysis techniques, however, are unable to efficiently unveil the underlying behavioural patterns and temporal dependencies necessary for effective anomaly detection. This work proposes a machine learning-driven framework for unveiling abstract behavioural patterns from SoC runtime activity logs with the synergistic combination of temporal graph embeddings and clustering-based abstraction and anomaly detection procedures. The scheme was tested on a RISC-V-based SoC prototype, using runtime traces under nominal and fault-injected scenarios. Experimental results demonstrate that the proposed approach enhances clustering quality, achieving an Adjusted Rand Index of 0.86 and a Normalized Mutual Information of 0.87, surpassing state-of-the-art baselines such as LogUAD and Log2graphs. Anomaly detection was obtained by the model with an F1-score of 0.91 and an AUC of 0.95, and evidence of stability in detecting deviations at low false alarm rates. Computational efficiency analysis also indicates that inference latency reduces by ~26% compared to graph-based baselines, with the ability to support real-time monitoring. These results establish that the intended approach not only enables more design-time verification of SoC systems but also facilitates secure runtime fault monitoring. The paper concludes that machine learning-based behavioural abstraction is an operationally tractable, interpretable, and scalable solution for enhancing SoC dependability. Subsequent research will deploy the approach to heterogeneous log sources, energy-constrained optimization, and adaptive online learning across changing workloads.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.005 | 0.002 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.000 | 0.002 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".