RL-Enhanced LLMs and Rechargeable Jamming Mines: Achieving Zero-Trust Security for Hierarchical Drone Swarms
Bibliographic record
Abstract
Abstract This proposal introduces an innovative approach to achieving zero-trust anti-eavesdropping and anti-jamming capabilities in hierarchical drone swarms by employing reinforcement learning (RL) for boosting the potential of telecom-tuned on-device large language models (LLMs). We propose fine-tuning of on-device LLMs based on multi-modal data, including telecom-specific optimization algorithms for autonomous interference/jamming mitigation and anti-eavesdropping measures. To address the lack of emergent behavior in LLMs considering distributed collective intelligence paradigm, we propose a multi-agent RL (MARL) based approach to adaptively optimize policies based on real-time interactions. By treating the swarm of UXVs (i.e. drones) as ultra-dense networks (UDNs) with heterogeneous nodes such as fog or edge drones, we leverage MARL to exploit the inherent interference through optimum node association, resource block selection, and beam/power adjustment at each node. This efficiently mitigates interference at legitimate nodes while enhancing it elsewhere, adhering to a zero-trust approach by assuming eavesdroppers can be located anywhere. We also introduce the concept of rechargeable jamming mines (RJMs) onboard daughter or multi-role (edge) drones, which harvest energy from ambient radio frequency in the UDN environment. These RJMs are deployed and activated at strategic locations to create secure zones, maximizing the secure area around the drones by generating interference that disadvantages eavesdroppers, even with superior channel conditions. Our proposed approach has proven effective in typical UDN scenarios, where the MARL approach is used to optimize configuration settings at each base station, significantly enhancing secure area coverage. Furthermore, the solution is extendable for autonomous anti-jamming capabilities, allowing dynamic channel switching or transmission rate adaptation to mitigate jamming or interference. Our work aligns with the GENZERO24 vision of autonomous secure drone communication, offering robust protection against evolving threats and enhancing operational integrity in complex dynamic environments.
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 | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| 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".