Distributed IoT systems, such as smart homes or water distribution networks, support tasks such as monitoring, detection, actuation, and control. Yet their communication and management architectures remain information-agnostic. Sensors generate observations, protocols transform them into packets, routing and scheduling mechanisms optimize network-centric metrics such as connectivity, latency, and energy, and applications reconstruct decisions only after data has traversed the stack. This creates a mismatch between the information generated by the system and the information actually needed for decision-making: redundant or predictable data and signalling may be transmitted repeatedly, while observations essential for inference, coverage, control, or safety may be delayed, discarded, or under-prioritized. The project addresses this mismatch by developing an information-driven framework, grounded in traffic traces and datasets from real IoT networks, where sensing, communication, and network operation are guided by the structure and relevance of information. The goal is to move beyond packet-driven operation toward systems that understand what information is generated, how it relates to other observations, and how it should be communicated to support decisions. The framework operates across three connected levels, starting from measurable information descriptors. At the stack level, the project will analyze real IoT traffic across protocol layers, including application data, protocol signalling, and temporal patterns, to quantify entropy, predictability, novelty, and redundancy. These metrics will characterize when devices generate genuinely new information, when transmissions are predictable or redundant, and what residual information should still be communicated. At the physical/link level, communication signals will be treated not only as packet carriers, but also as sources of opportunistic sensing and compact information representation. Link-derived features such as RSSI, SNR, LQI, and temporal dynamics will provide additional observations of the environment and infrastructure state, while a limited set of robust signal primitives will test the minimum signalling needed for source identity, state changes, event flags or actuation triggers. At the system level, stack-level descriptors and link-derived observations will be aggregated to model how distributed observations relate to each other. Graph-based and Topological Signal Processing tools will be used to identify redundancy, complementarity, inconsistencies, structurally relevant sources, and persistent informational gaps. These system-level indicators will then support information-driven network decisions. Expected outcomes include a topology-driven view of distributed IoT systems, information-aware mechanisms reducing redundant traffic, improved robustness under realistic and degraded conditions, and a concrete pathway toward task-aware communication in real-world IoT networks.