UniShield: Unified Face Attack Detection via KG-Informed Multimodal Reasoning

TL;DR

UniShield introduces a knowledge-grounded multimodal reasoning framework for unified face attack detection, leveraging a structured attack knowledge graph and instruction tuning to enhance detection accuracy and reasoning reliability.

Contribution

The paper presents a novel framework combining a face attack knowledge graph with multimodal reasoning and a graph-consistent optimization to improve face attack detection.

Findings

Achieves high accuracy and low error rates on a new multimodal UAD benchmark.

Effectively links attack categories to visual cues for better reasoning.

Outperforms baseline methods in detection and explanation quality.

Abstract

Unified face attack detection (UAD) requires recognizing physical spoofing and digital forgery within a shared decision space, yet existing discriminative or prompt-based methods largely rely on appearance correlations and provide limited evidence-grounded reasoning. We propose UniShield, a knowledge-grounded multimodal reasoning framework for unified face attack defense. UniShield constructs a Face Attack Knowledge Graph (FAKG) that links attack categories to diagnostic visual cues and attack-conditioned relations, and uses it to synthesize 52,025 FAKG-QA examples for Attack-Graph Instruction Tuning (AGIT). To improve rationale consistency, we further introduce Graph-Consistent Reasoning Optimization (GCRO), a GRPO-based objective with a KG-consistency reward that encourages generated rationales to match graph-supported cues while penalizing incompatible claims. Experiments on our multimodal UAD benchmark show that UniShield achieves strong performance across binary, coarse-grained, and fine-grained protocols, with consistently high ACC and low HTER. These results suggest that structured attack knowledge can improve both detection accuracy and reasoning reliability over discriminative baselines and general-purpose MLLMs. Our code will be released at https://anonymous.4open.science/r/Unishield-A6A3/.