Deep Reinforcement Learning for Penetration Testing of Cyber-Physical Attacks in the Smart Grid
Bibliographic record
Abstract
The fast expansion of interconnectivity in cyber-physical critical infrastructures like smart grids has given rise to concerning exposures and vulnerabilities. Although penetration testing (PT) has been an effective approach to searching for vulnerabilities in software, devices, and networks from the attacker's view, the strong cyber-physical coupling in these large-scale infrastructures has made it challenging to manually pinpoint critical vulnerabilities, particularly at system levels due to the complexity, dimensionality, and uncertainty therein. To better protect the security of cyber-physical systems, this paper proposes a deep reinforcement learning (DRL)-based PT framework to efficiently and adaptively identify critical vulnerabilities in smart grids. Using replay attacks as an example, the paper models the attack as a Markov Decision Process with three actions - stop, record, and replay - to learn the optimal timing and ordering of replays in different operating scenarios. A cyber-physical co-simulation platform with dedicated simulators for the physical part, cyber part, control part, and attacker part of a smart distribution grid was developed as a sandbox environment to train the DRL agent. Scenarios with different levels of difficulty are tested to validate the learning capability and performance in finding critical attack paths of the DRL-based PT. The simulation results show that DRL-based PT can learn to find the optimal attack path against system stability when the grid is under high load demand, solar power generation, and weather variation. These results are promising first steps toward a highly customizable framework to pen-test complex cyber-physical systems with automatic DRL agents and various attack schemes.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| 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".