Evidence · DiggingBeagle record
MobileProbe selects reachable UI injection points before evaluating mobile agents (research paper)
MobileProbe shows that placing hostile UI content where a mobile agent will encounter it changes measured attack success, even before the payload is optimized.
- Published
- Nov 12, 2026
- Source role
- primary disclosure
Evidence record
MobileProbe shows that placing hostile UI content where a mobile agent will encounter it changes measured attack success, even before the payload is optimized.
Claim-level citations (4)
- supportsMobileProbe selects reachable UI injection points before evaluating mobile agents: The 83 MobileProbe attack cases have two materially different attacker prerequisites: 44 use malicious user-generated material in otherwise available apps, while 39 use attacker-repackaged applications to display fabricated interface elements. The benchmark verifies selected actions and device states after execution, not just whether an agent repeated the attacker text.
Dataset and evaluation methodology, distribution of 44 user-generated-content and 39 repackaged-app attacks
- supportsMobileProbe selects reachable UI injection points before evaluating mobile agents: The paper reports 85.1, 86.3 and 85.1 percent completion of selected attacker objectives for three evaluated agents, materially above its specified static-template and zero-shot controls.
Results, overall attack-success comparison and ablations
- supportsMobileProbe selects reachable UI injection points before evaluating mobile agents: Researchers observe a benign GUI trajectory, select editable text or fabricated-dialog surfaces near a real decision, generate and refine hostile instructions and then measure the simulated Android action rather than a verbal promise.
Framework, trajectory selection and payload evaluation stages
- supportsMobileProbe selects reachable UI injection points before evaluating mobile agents: The paper reports that removing evolutionary payload refinement reduces Gemini attack success from 85.1 percent to 28.5 percent; trajectory-aware placement and execution feedback are material parts of this selected benchmark, not incidental prompt wording.
Section VI, Table VI component ablations on Gemini-3.1-Pro