Towards Secure Tuning: Mitigating Security Risks Arising from Benign Instruction Fine-Tuning

Instruction Fine-Tuning (IFT) has become an essential method for adapting base Large Language Models (LLMs) into variants for professional and private use. However, researchers have raised concerns over a significant decrease in LLMs' security following IFT, even when the IFT process involves entirely benign instructions (termed Benign IFT). Our study represents a pioneering effort to mitigate the security risks arising from Benign IFT. Specifically, we conduct a Module Robustness Analysis, aiming to investigate how LLMs' internal modules contribute to their security. Based on our analysis, we propose a novel IFT strategy, called the Modular Layer-wise Learning Rate (ML-LR) strategy. In our analysis, we implement a simple security feature classifier that serves as a proxy to measure the robustness of modules (e.g. $Q$/$K$/$V$, etc.). Our findings reveal that the module robustness shows clear patterns, varying regularly with the module type and the layer depth. Leveraging these insights, we develop a proxy-guided search algorithm to identify a robust subset of modules, termed Mods$_{Robust}$. During IFT, the ML-LR strategy employs differentiated learning rates for Mods$_{Robust}$ and the rest modules. Our experimental results show that in security assessments, the application of our ML-LR strategy significantly mitigates the rise in harmfulness of LLMs following Benign IFT. Notably, our ML-LR strategy has little impact on the usability or expertise of LLMs following Benign IFT. Furthermore, we have conducted comprehensive analyses to verify the soundness and flexibility of our ML-LR strategy.

Dynamic Decentralized 3D Urban Coverage and Patrol with UAVs

In the event of natural or man-made disasters in an urban environment, such as fires, floods, and earthquakes, a swarm of unmanned aerial vehicles (UAVs) can rapidly sweep and provide coverage to monitor the area of interest and locate survivors. We propose a modular framework and patrol strategy that enables a swarm of UAVs to perform cooperative and periodic coverage in such scenarios. Our approach first discretizes the area of interest into viewpoints connected via closed paths. UAVs are assigned to teams via task allocation to cooperatively patrol these closed paths. We propose a minimal, scalable, and robust patrol strategy where UAVs within a team move in a random direction along their assigned closed path and "bounce" off each other when they meet. Our simulation results show that such a minimal strategy can exhibit an emergent behaviour that provides periodic and complete coverage in a 3D urban environment.