Alina Oprea

The paper introduces a Retrieval-Augmented Generation (RAG) system that uses targeted query filtering and LLM semantic reasoning to accurately and cost-effectively analyze complex cybersecurity incidents from diverse log sources.

The paper introduces BOA, a novel framework that measures agent safety by exhaustively searching the entire in-budget trajectory space, thereby identifying unsafe behaviors missed by traditional sampling methods.

The paper introduces MAGIQ, a novel, quantum-resistant framework designed to securely define and enforce communication and access-control policies within multi-agent AI systems.

This paper analyzes attacks against centralized agent governance systems (SAGA) when the central provider is compromised and proposes three novel, trade-off-aware architectures (SAGA-BFT, SAGA-MON, SAGA-AUD, SAGA-HYB) to enhance Byzantine resilience.

This paper investigates the privacy risk of reconstructing Personally Identifiable Information (PII) from Large Language Models (LLMs) that have undergone Supervised Finetuning (SFT), proposing a novel decoding algorithm (COVA) for defense evaluation.

The paper addresses the 'agent attribution' problem—the inability to trace harmful or misbehaving AI agents back to their deploying account—by proposing a robust, canary-based protocol for vendors to identify the responsible user.

The paper introduces PoisonForge, a comprehensive benchmark demonstrating that even a small number of targeted poisoned examples can significantly compromise the safety and reliability of instruction-tuned LLMs across various model sizes.

This paper proposes a Bayesian framework to enhance membership inference attacks against released statistics by incorporating prior knowledge about the population's attribute dependency structure, outperforming existing methods.