Mobile agent auditor detects security rule violations with high accuracy
MATE: Policy-Aware Security Auditing for Mobile Agents via Synthesis-Driven Trajectory Learning
Cryptography and SecurityArtificial Intelligence
Summary
Mobile agents help automate tasks on smartphones but can accidentally break security rules unique to each app. The authors created MATE, a tool that understands both these rules written in everyday language and the mobile agent’s actions to spot when rules are broken and explain why. Instead of hardcoding rules, MATE lets users change them easily without needing to guess how the underlying AI works. They trained MATE using a large set of simulated mobile app actions and showed it works very well on real and synthetic tests, catching problems better than earlier tools.
What this means in practice
- •For mobile app developers: Audit complex mobile agent actions automatically to check compliance with app-specific security rules written in natural language.
- •For enterprise security teams: Improve detection of policy violations by autonomous software agents running on mobile devices to reduce security risks.
Authors
Changyue Jiang, Jiayi Wang, Xin Wen, Jiarun Dai, Geng Hong, Xudong Pan
Abstract
Mobile agents powered by foundation models now automate complex, multi-step workflows on real devices, but their trajectories can violate app-specific security policies. Existing trajectory-level defenses rely on LLM prompting or rigid rules, and thus fail to support fine-grained, natural-language policies that generalize across apps and tasks. In this work, we introduce MATE, a lightweight, policy-conditioned auditor that encodes both agent trajectories and natural-language security policies to determine whether a trajectory violates a given policy and to explain why. Treating policies as editable text rather than fixed model parameters allows MATE to handle user-defined and evolving requirements without retraining. To construct MATE, we build a knowledge base by extracting app descriptions, workflows, and policies from hundreds of popular mobile apps worldwide, and synthesizing over 140K semantically realistic, policy-conditioned trajectories with a multi-stage pipeline. We further release MATEBench, a trajectory-level auditing benchmark with two synthetic subsets and one real-world subset of manually collected trajectories. Models trained with our synthesis-driven trajectory learning achieve over 95% accuracy on MATEBench, retain strong performance on external safety benchmarks, and audit trajectories from Zhipu's AutoGLM and Alibaba's Mobile-Agent on real devices with over 95% accuracy, outperforming prior methods by over 20%. MATE shows that practical, fine-grained security auditing for heterogeneous mobile agents is both feasible and effective.