DIBench: Benchmarking Decision Integrity of GUI-based Mobile Agents Under Deceptive Injections
cs.CR, cs.MA
Submitted: 2026-09-29
Updated: 2026-09-29
Comments: Accepted to NeurIPS 2026, Track on Evaluations and Datasets
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- What Is Your AI Agent Buying? Evaluation, Biases, Model Dependence, & Emerging Implications for Agentic E-Commerce
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- The Obvious Invisible Threat: LLM-Powered GUI Agents' Vulnerability to Fine-Print Injections
- Toward a Human-Centered Evaluation Framework for Trustworthy LLM-Powered GUI Agents
- GhostEI-Bench: Do Mobile Agents Resilience to Environmental Injection in Dynamic On-Device Environments?
- D-GARA: A Dynamic Benchmarking Framework for GUI Agent Robustness in Real-World Anomalies
- Measuring the Security of Mobile LLM Agents under Adversarial Prompts from Untrusted Third-Party Channels
- A uniform bound on the smallest surjective prime of an elliptic curve
- Tool Preferences in Agentic LLMs are Unreliable
- The Behavioral Fabric of LLM-Powered GUI Agents: Human Values and Interaction Outcomes
- Countering Adversarial Images using Input Transformations
- A Critical Evaluation of Defenses against Prompt Injection Attacks
- Cross-Modal Content Optimization for Steering Web Agent Preferences
- TRAP: Targeted Redirecting of Agentic Preferences
- Aligning Language Models to Explicitly Handle Ambiguity
- MobileWorldBench: Towards Semantic World Modeling For Mobile Agents
- AppAgent v2: Advanced Agent for Flexible Mobile Interactions
- Mind the Third Eye! Benchmarking Privacy Awareness in MLLM-powered Smartphone Agents
- LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs