Stealthy Black-Box Attacks on Deep Learning Non-Intrusive Load Monitoring Models
Bibliographic record
Abstract
With the advent of the advanced metering infrastructure, electricity usage data is being continuously generated at large volumes by smart meters vastly deployed across the modern power grid. Electric power utility companies and third party entities such as smart home management solution providers gain significant insights into these datasets via machine learning (ML) models. These are then utilized to perform active/passive power demand management that fosters economical and sustainable electricity usage. Although ML models are powerful, these remain vulnerable to adversarial attacks. A novel stealthy black-box attack construction model is proposed that targets deep learning models utilized to perform non-intrusive load monitoring based on smart meter data. These attacks are practical as there is no assumption of the knowledge of training data, internal parameters, and architecture of the targeted ML model. The profound impact of the proposed stealthy attack constructions on energy analytics and decision-making processes is shown through comprehensive theoretical, practical, and comparative analysis. This work sheds light on vulnerabilities of ML models in the smart grid context and provides valuable insights for securely accommodating increasing prevalence of artificial intelligence in the modern power grid.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 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 source (direct Gemma or distilled Codex), 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".