Controlled Evaluation of HID/USB and Ethernet Drop-Device Attacks in Endpoint and SDN Environments
Year: 2026
Type: conference-paper
Abstract: SenseAR is an assistive reading prototype that captures printed material, performs optical character recognition, and returns synthesized speech, with processing that may run on a mobile device, a local server, or a cloud environment. This flexibility widens the attack surface: images of documents, recognized text, and generated audio traverse device, network, and external-service boundaries. This paper presents a cybersecurity architecture for such an edge-cloud assistive pipeline. We build a data-flow threat model across the capture, pre-processing, OCR, text-preparation, and text-to-speech stages, map trust boundaries, and derive protection requirements centered on data minimization, authenticated and encrypted transport, least-privilege local execution, and controlled retention of raw material. Drawing on a companion penetration-testing study of an embedded network gateway on the same laboratory network, we show how a weak boundary device (deprecated TLS without forward secrecy, single weak SSH key-exchange, plaintext admin authentication) undermines the confidentiality of an otherwise locally executed pipeline, and we translate those findings into concrete controls for SenseAR. The result is a layered defense that keeps captured personal documents confidential while preserving the modular capture-OCR-TTS structure.
Institution: Technical University of Sofia
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Funder: Technical University of Sofia
OpenAlex ID: W7225753352