Abstract

Recent LLM-based systems have shown promising capabilities for security-focused code analysis, including vulnerability identification and reverse engineering. Malware understanding, however, poses a distinct challenge: analysts must reconstruct high-level malicious behaviors under partial observability from sparse, dispersed evidence intertwined with benign functionality. While static analysis can expose security-relevant signals, the central challenge is not merely identifying suspicious code, but determining whether the evidence sufficiently supports an auditable behavior-level conclusion. We formulate malware understanding as a grounded reasoning problem and argue that reliable behavior reconstruction requires three complementary forms of grounding. Domain grounding constrains how behavior hypotheses are generated and evaluated, semantics grounding localizes and connects supporting program evidence, and knowledge grounding supports behavioral attribution through externally verifiable threat knowledge. To study this hypothesis, we present Malaika, a tri-grounded multi-agent framework that operationalizes the three grounding mechanisms through analyst-inspired reasoning, tool-mediated evidence localization, explicit review, and retrieval-based behavioral attribution. We instantiate Malaika for Android malware analysis and evaluate it on malware-understanding tasks with analyst-validated behaviorlevel annotations. Our results show that Malaika improves analysis quality over prior LLM-based malware-analysis frameworks and demonstrate that reliability depends not only on model capability but also on the structure of the reasoning process itself. In particular, comparisons against both malware-analysis systems and frontier agentic frameworks show that behavioral attribution is strongly influenced by the surrounding reasoning process, with grounding-aware reasoning producing substantially more precise and auditable conclusions. More importantly, ablation studies support the grounding hypothesis: domain grounding, semantics grounding, and knowledge grounding address complementary failure modes by contributing distinct capabilities for hypothesis generation, evidence localization, and behavioral attribution, respectively. These findings suggest that grounding-aware reasoning provides a principled foundation for reliable malware understanding and, more broadly, for evidence-grounded software analysis.