Retrieval-Augmented LLMs for Security Incident Analysis

Investigating cybersecurity incidents requires collecting and analyzing evidence from multiple log sources, including intrusion detection alerts, network traffic records, and authentication events. This process is labor-intensive: analysts must sift through large volumes of data to identify relevant indicators and piece together what happened. We present a RAG-based system that performs security incident analysis through targeted query-based filtering and LLM semantic reasoning. The system uses a query library with associated MITRE ATT\&CK techniques to extract indicators from raw logs, then retrieves relevant context to answer forensic questions and reconstruct attack sequences. We evaluate the system with five LLM providers on malware traffic incidents and multi-stage Active Directory attacks. We find that LLM models have different performance and tradeoffs, with Claude Sonnet~4 and DeepSeek~V3 achieving 100\% recall across all four malware scenarios, while DeepSeek costs 15$\times$ less (\$0.008 vs.\ \$0.12 per analysis). Attack step detection on Active Directory scenarios reaches 100\% precision and 82\% recall. Ablation studies confirm that a RAG architecture is essential: LLM baselines without RAG-enhanced context correctly identify victim hosts but miss all attack infrastructure including malicious domains and command-and-control servers. These results demonstrate that combining targeted query-based filtering with RAG-based retrieval enables accurate, cost-effective security analysis within LLM context limits.

ABCD: All Biases Come Disguised

Multiple-choice question (MCQ) benchmarks have been a standard evaluation practice for measuring LLMs' ability to reason and answer knowledge-based questions. Through a synthetic NonsenseQA benchmark, we observe that different LLMs exhibit varying degrees of label-position-few-shot-prompt bias, where the model either uses the answer position, the label in front of the answer, the distributions of correct answers present in the few-shot prompt, or a combination of all to answer each MCQ question. We propose a simple bias-reduced evaluation protocol that replaces the labels of each question with uniform, unordered labels and prompts the LLM to use the whole answer presented. With a simple sentence similarity model, we demonstrate improved robustness and lower standard deviation between different permutations of answers with a minimal drop in LLM's performance, exposing the LLM's capabilities under reduced evaluation artifacts, without any help from the prompt examples or the option labels. Across multiple benchmarks and models, this protocol substantially improves the robustness to answer permutations, reducing mean accuracy variance $3\times$ with only a minimal decrease in the mean model's performance. Through ablation studies on various embedding models and similarity functions, we show that the method is more robust than the standard ones.

Simplify to Amplify: Achieving Information-Theoretic Bounds with Fewer Steps in Spectral Community Detection

We propose a streamlined spectral algorithm for community detection in the two-community stochastic block model (SBM) under constant edge density assumptions. By reducing algorithmic complexity through the elimination of non-essential preprocessing steps, our method directly leverages the spectral properties of the adjacency matrix. We demonstrate that our algorithm exploits specific characteristics of the second eigenvalue to achieve improved error bounds that approach information-theoretic limits, representing a significant improvement over existing methods. Theoretical analysis establishes that our error rates are tighter than previously reported bounds in the literature. Comprehensive experimental validation confirms our theoretical findings and demonstrates the practical effectiveness of the simplified approach. Our results suggest that algorithmic simplification, rather than increasing complexity, can lead to both computational efficiency and enhanced performance in spectral community detection.

PoolFlip: A Multi-Agent Reinforcement Learning Security Environment for Cyber Defense

Cyber defense requires automating defensive decision-making under stealthy, deceptive, and continuously evolving adversarial strategies. The FlipIt game provides a foundational framework for modeling interactions between a defender and an advanced adversary that compromises a system without being immediately detected. In FlipIt, the attacker and defender compete to control a shared resource by performing a Flip action and paying a cost. However, the existing FlipIt frameworks rely on a small number of heuristics or specialized learning techniques, which can lead to brittleness and the inability to adapt to new attacks. To address these limitations, we introduce PoolFlip, a multi-agent gym environment that extends the FlipIt game to allow efficient learning for attackers and defenders. Furthermore, we propose Flip-PSRO, a multi-agent reinforcement learning (MARL) approach that leverages population-based training to train defender agents equipped to generalize against a range of unknown, potentially adaptive opponents. Our empirical results suggest that Flip-PSRO defenders are $2\times$ more effective than baselines to generalize to a heuristic attack not exposed in training. In addition, our newly designed ownership-based utility functions ensure that Flip-PSRO defenders maintain a high level of control while optimizing performance.

MA-RAG: Multi-Agent Retrieval-Augmented Generation via Collaborative Chain-of-Thought Reasoning

We present MA-RAG, a Multi-Agent framework for Retrieval-Augmented Generation (RAG) that addresses the inherent ambiguities and reasoning challenges in complex information-seeking tasks. Unlike conventional RAG methods that rely on either end-to-end fine-tuning or isolated component enhancements, MA-RAG orchestrates a collaborative set of specialized AI agents: Planner, Step Definer, Extractor, and QA Agents, to tackle each stage of the RAG pipeline with task-aware reasoning. Ambiguities may arise from underspecified queries, sparse or indirect evidence in retrieved documents, or the need to integrate information scattered across multiple sources. MA-RAG mitigates these challenges by decomposing the problem into subtasks, such as query disambiguation, evidence extraction, and answer synthesis, and dispatching them to dedicated agents equipped with chain-of-thought prompting. These agents communicate intermediate reasoning and progressively refine the retrieval and synthesis process. Our design allows fine-grained control over information flow without any model fine-tuning. Crucially, agents are invoked on demand, enabling a dynamic and efficient workflow that avoids unnecessary computation. This modular and reasoning-driven architecture enables MA-RAG to deliver robust, interpretable results. Experiments on multi-hop and ambiguous QA benchmarks demonstrate that MA-RAG outperforms state-of-the-art training-free baselines and rivals fine-tuned systems, validating the effectiveness of collaborative agent-based reasoning in RAG.

Quantitative Resilience Modeling for Autonomous Cyber Defense

Cyber resilience is the ability of a system to recover from an attack with minimal impact on system operations. However, characterizing a network's resilience under a cyber attack is challenging, as there are no formal definitions of resilience applicable to diverse network topologies and attack patterns. In this work, we propose a quantifiable formulation of resilience that considers multiple defender operational goals, the criticality of various network resources for daily operations, and provides interpretability to security operators about their system's resilience under attack. We evaluate our approach within the CybORG environment, a reinforcement learning (RL) framework for autonomous cyber defense, analyzing trade-offs between resilience, costs, and prioritization of operational goals. Furthermore, we introduce methods to aggregate resilience metrics across time-variable attack patterns and multiple network topologies, comprehensively characterizing system resilience. Using insights gained from our resilience metrics, we design RL autonomous defensive agents and compare them against several heuristic baselines, showing that proactive network hardening techniques and prompt recovery of compromised machines are critical for effective cyber defenses.

VoD-3DGS: View-opacity-Dependent 3D Gaussian Splatting

Reconstructing a 3D scene from images is challenging due to the different ways light interacts with surfaces depending on the viewer's position and the surface's material. In classical computer graphics, materials can be classified as diffuse or specular, interacting with light differently. The standard 3D Gaussian Splatting model struggles to represent view-dependent content, since it cannot differentiate an object within the scene from the light interacting with its specular surfaces, which produce highlights or reflections. In this paper, we propose to extend the 3D Gaussian Splatting model by introducing an additional symmetric matrix to enhance the opacity representation of each 3D Gaussian. This improvement allows certain Gaussians to be suppressed based on the viewer's perspective, resulting in a more accurate representation of view-dependent reflections and specular highlights without compromising the scene's integrity. By allowing the opacity to be view dependent, our enhanced model achieves state-of-the-art performance on Mip-Nerf, Tanks&Temples, Deep Blending, and Nerf-Synthetic datasets without a significant loss in rendering speed, achieving >60FPS, and only incurring a minimal increase in memory used.

Teaching Wav2Vec2 the Language of the Brain

The decoding of continuously spoken speech from neuronal activity has the potential to become an important clinical solution for paralyzed patients. Deep Learning Brain Computer Interfaces (BCIs) have recently successfully mapped neuronal activity to text contents in subjects who attempted to formulate speech. However, only small BCI datasets are available. In contrast, labeled data and pre-trained models for the closely related task of speech recognition from audio are widely available. One such model is Wav2Vec2 which has been trained in a self-supervised fashion to create meaningful representations of speech audio data. In this study, we show that patterns learned by Wav2Vec2 are transferable to brain data. Specifically, we replace its audio feature extractor with an untrained Brain Feature Extractor (BFE) model. We then execute full fine-tuning with pre-trained weights for Wav2Vec2, training ''from scratch'' without pre-trained weights as well as freezing a pre-trained Wav2Vec2 and training only the BFE each for 45 different BFE architectures. Across these experiments, the best run is from full fine-tuning with pre-trained weights, achieving a Character Error Rate (CER) of 18.54\%, outperforming the best training from scratch run by 20.46\% and that of frozen Wav2Vec2 training by 15.92\% percentage points. These results indicate that knowledge transfer from audio speech recognition to brain decoding is possible and significantly improves brain decoding performance for the same architectures. Related source code is available at https://github.com/tfiedlerdev/Wav2Vec2ForBrain.

Explore Reinforced: Equilibrium Approximation with Reinforcement Learning

Current approximate Coarse Correlated Equilibria (CCE) algorithms struggle with equilibrium approximation for games in large stochastic environments but are theoretically guaranteed to converge to a strong solution concept. In contrast, modern Reinforcement Learning (RL) algorithms provide faster training yet yield weaker solutions. We introduce Exp3-IXrl - a blend of RL and game-theoretic approach, separating the RL agent's action selection from the equilibrium computation while preserving the integrity of the learning process. We demonstrate that our algorithm expands the application of equilibrium approximation algorithms to new environments. Specifically, we show the improved performance in a complex and adversarial cybersecurity network environment - the Cyber Operations Research Gym - and in the classical multi-armed bandit settings.

ZipNN: Lossless Compression for AI Models

With the growth of model sizes and the scale of their deployment, their sheer size burdens the infrastructure requiring more network and more storage to accommodate these. While there is a vast model compression literature deleting parts of the model weights for faster inference, we investigate a more traditional type of compression - one that represents the model in a compact form and is coupled with a decompression algorithm that returns it to its original form and size - namely lossless compression. We present ZipNN a lossless compression tailored to neural networks. Somewhat surprisingly, we show that specific lossless compression can gain significant network and storage reduction on popular models, often saving 33% and at times reducing over 50% of the model size. We investigate the source of model compressibility and introduce specialized compression variants tailored for models that further increase the effectiveness of compression. On popular models (e.g. Llama 3) ZipNN shows space savings that are over 17% better than vanilla compression while also improving compression and decompression speeds by 62%. We estimate that these methods could save over an ExaByte per month of network traffic downloaded from a large model hub like Hugging Face.