Analysis of the Use of Artificial Intelligence Language Models in Mobile Malware (Case Study of Modern Malware PromptSpy)

Objective. The study aims to analyze the use of artificial intelligence language models within next-generation mobile malware - specifically the modern malicious software known as "PromptSpy" - by examining its operational mechanisms and the role of its AI module as a critical component in the execution of an attack. Method. A structural-functional analysis of the publicly available technical documentation for PromptSpy was conducted. The identified features were cross-referenced with official Android operating system documentation, including specifications for the `AccessibilityService` component and the system’s "Recents" screen interface. Result. PromptSpy should be viewed as an early iteration of Android malware that employs large language models (LLMs) not to replace its entire malicious feature set, but to enhance the most labor-intensive aspect of hard-coded automation: adapting to a device’s user interface. The study reconstructs the malware’s sequence of operations: PromptSpy captures an XML snapshot of the current user interface state, transmits it to the Gemini model, receives structured instructions for the next action, and repeats this cycle until the interface reaches the state required to proceed with further malicious activity. Conclusion. The integration of a generative AI model into mobile malware demonstrates a shift from rigidly programmed UI interaction scripts to context-adaptive malicious logic. The PromptSpy case shows that even local integration of AI into a single functional node significantly increases the threat’s adaptability and complicates detection, as variability moves from the level of packaging and obfuscation to the level of environment interpretation and action selection during execution.

Authors

Institutions

Publication Details

Journal
Herald of Dagestan State Technical University Technical Sciences
Published
2026-11-12
DOI
https://doi.org/10.21822/2073-6185-2026-53-3-89-95
Primary Topic
Advanced Malware Detection Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Analysis of the Use of Artificial Intelligence Language Models in Mobile Malware (Case Study of Modern Malware PromptSpy)

О. В. Ланкин, Е. А. Рогозин, Aleksey Lukoyanov, R. E. Rogozin
Herald of Dagestan State Technical University Technical Sciences
Advanced Malware Detection Techniques
article

Analysis of the Use of Artificial Intelligence Language Models in Mobile Malware (Case Study of Modern Malware PromptSpy)

О. В. Ланкин, Е. А. Рогозин, Aleksey Lukoyanov, R. E. Rogozin
article en

Abstract

Objective. The study aims to analyze the use of artificial intelligence language models within next-generation mobile malware - specifically the modern malicious software known as "PromptSpy" - by examining its operational mechanisms and the role of its AI module as a critical component in the execution of an attack. Method. A structural-functional analysis of the publicly available technical documentation for PromptSpy was conducted. The identified features were cross-referenced with official Android operating system documentation, including specifications for the `AccessibilityService` component and the system’s "Recents" screen interface. Result. PromptSpy should be viewed as an early iteration of Android malware that employs large language models (LLMs) not to replace its entire malicious feature set, but to enhance the most labor-intensive aspect of hard-coded automation: adapting to a device’s user interface. The study reconstructs the malware’s sequence of operations: PromptSpy captures an XML snapshot of the current user interface state, transmits it to the Gemini model, receives structured instructions for the next action, and repeats this cycle until the interface reaches the state required to proceed with further malicious activity. Conclusion. The integration of a generative AI model into mobile malware demonstrates a shift from rigidly programmed UI interaction scripts to context-adaptive malicious logic. The PromptSpy case shows that even local integration of AI into a single functional node significantly increases the threat’s adaptability and complicates detection, as variability moves from the level of packaging and obfuscation to the level of environment interpretation and action selection during execution.

Herald of Dagestan State Technical University Technical SciencesVol. 53(3)
Voronezh Institute of Ministry of Internal Affairs (RU)
Advanced Malware Detection Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon