A Novel Privacy-Preserving Framework For Mobile Malware Detection: Design and Threat Modelling

Without a doubt, mobile devices have become primary endpoints for communication, work, and personal life. Due to this popularity, cyberattacks on mobile de-vices are becoming more sophisticated, necessitating strong detection methods that often require deep access to sensitive user data. However, current detection methods typically lack crucial user-rich behaviour data, which is vital for training effective models due to stringent privacy-preserving policies. To resolve this con-flict, the paper proposes a theoretical framework for privacy-preserving malware detection that combines (I) cryptographic primitives such as symmetric encryption (like Advanced Encryption Standard (AES) for data at rest), asymmetric encryption (like Elliptic Curve Cryptography (ECC) for key transport), and option-ally homomorphic encryption (HE) for analysis; (II) anonymity protocols like network-level onion-routing-inspired relay and data-level k-anonymity; and (III) mathematical models such as entropy analysis, Markov chains, and graph-based reasoning to enable malware detection without revealing user identities. We formally define a "Privacy-Preserving Malware Detection Engine" and introduce a rigorous threat model based on the "honest-but-curious" server assumption. We then outline design principles and analyse security and privacy properties—confidentiality, integrity, anonymity, unlinkability, and utility trade-offs. This theoretical foundation clarifies assumptions and guarantees, providing a blueprint for future implementations and evaluations in mobile environments.