Abdullahil Kafi | Search arXiv

arXiv2026

GRIDSHIELD: Verifiable Grid Feasibility and Manipulation Resistance in Privacy-Preserving Peer-to-Peer Energy Trading

Blockchain-based peer-to-peer (P2P) energy markets let neighbors trade directly, but concentrate unchecked power in the clearing operator. Most prior work addresses only part of this: network-aware clearing respects grid constraints but operates on plaintext allocations, while privacy-preserving designs hide bids yet either drop the network constraint, enforce it only inside an unauditable computation, or recover it at the cost of a per-circuit trusted setup. Neither considers what happens if an adversary captures the operator's clearing privilege to reorder trades for profit or curtail a rival. We present GRIDSHIELD, a P2P energy-trading system that clears sealed bids with an order-independent batch program, removing the welfare a compromised operator can extract by reordering trades, and proves the result grid-feasible via an aggregated congestion-feasibility proof (ACFP). Exploiting the additive homomorphism of Pedersen commitments, ACFP reconstructs each grid constraint directly from per-bus commitments, reducing feeder-wide feasibility verification over any public grid constraints to a single aggregated range proof over the hidden allocation, with a reveal-and-recompute challenge that catches feasible yet unfair clearings. We introduce congestion-responsibility fairness (CRF), a metric scoring curtailment against each party's physical responsibility for congestion rather than consumption. We evaluate GRIDSHIELD against three prior P2P energy-trading designs across four feeders, including a 240-node Midwest feeder with real hourly loads, on a live Hyperledger Fabric network benchmarked with Hyperledger Caliper. GRIDSHIELD proves an entire round feasible with a few-kilobyte proof and no trusted setup, rejects infeasible allocations two of the three compared designs accept, and sustains $\approx$1064 tx/s.

cs.CR↗

arXiv2026

RAG-NAROK: Retrieval-Aware Knowledge Corpus Poisoning in RAG with Source-specific Refutation

Retrieval augmented generation (RAG) systems have emerged as the dominant architecture for grounding large language model (LLM) outputs in verifiable external knowledge, yet their structural reliance on a dynamic retrieval pipeline introduces a largely unexplored class of adversarial vulnerability. Existing knowledge-base poisoning attacks are fundamentally static. Adversarial documents are pre-computed and injected without any awareness of what the victim system will actually retrieve for a given query, leaving the attack blind to the competitive documentary landscape that surrounds its payload in the generator's context window. Unlike traditional static poisoning attacks that are blind to the retrieved context, we introduce RAG-NAROK (Retrieval-Anchored Generation Negation And Response Quality Collapse), a RAG attack framework that adapts to the query text. RAG-NAROK exploits the transparency inherent in RAG pipeline to first extract the legitimate source identities, then generate Anchor-Specific Refutation documents that explicitly name and devalue retrieved sources while leveraging recency and authority biases to steer the text generation toward a target answer. Our results demonstrate that RAG-NAROK significantly outperforms static baselines across diverse domains, revealing a fundamental tension between RAG transparency and AI security.

cs.AI↗