Abnormality Detection and Localization Schemes Using Molecular Communication Systems: A Survey

Abstract

Abnormality detection and localization (ADL) have been studied widely in wireless sensor networks (WSNs) literature, where the sensors use electromagnetic waves for communication. Molecular communication (MC) has been introduced as an alternative approach for ADL in particular areas such as healthcare, being able to tackle the shortcomings of conventional WSNs, such as invasiveness, bioincompatibility, and high energy consumption. In this paper, we introduce a general framework for MCbased ADL, which consists of multiple tiers for sensing the abnormality and communication between different agents, including the sensors, the fusion center (FC), the gateway (GW), and the external node (e.g., a local cloud), and describe each tier and the agents in this framework. We classify and explain different abnormality recognition methods, the functional units of the sensors, and different sensor features. Further, we describe different types of interfaces required for converting the internal and external signals at the FC and GW. Moreover, we present a unified channel model for the sensing and communication links. We categorize the MC-based abnormality detection schemes based on the sensor mobility, cooperative detection, and cooperative sensing/activation. We also classify the localization approaches based on the sensor mobility and propulsion mechanisms and present a general framework for the externally-controllable localization systems. Finally, we present some challenges and future research directions to realize and develop MC-based systems for ADL. The important challenges in the MC-based systems lie in four main directions as implementation, system design, modeling, and methods, which need considerable attention from multidisciplinary perspectives.

Hybrid Deep Learning Techniques for Securing Bioluminescent Interfaces in Internet of Bio Nano Things

“…Reducing human intervention during feature selection, the framework decreased system complexity for future adoption in 5G IoT networks to isolate traffic of interest. Etemadi et al [31] surveyed abnormality detection at the molecular communication level and compared different ML, heuristics, and DL primitives for this purpose. The work discussed the viability of ML approaches, as well as the utilization of naturally inspired systems based on genetic, particle swarm, and colony optimization algorithms for identifying anomalous (traffic) activity.…”

“…Post-transformation dataset, therefore, results in a greater number of features due to one hot encoding of traffic class labeling. Compared to other nominal conversion schemes such as Bag-of-Word (BoW), and MinMaxScaling, one hot encoding provides a single-stage conversion suitable for BBI parameters relying on substantially lower computing capability favored in IoT environments [29,31]. The pre-processed dataset is subsequently input to neural algorithms using control variables described in the next sub-section.…”

“…These early non-ML reliant schemes allowed limited toleration for noisy signals affecting in-vivo TDD, while also being computationally intensive due to inherent multi-stage encryptions. Among ML/DL classification use-cases in IoT, and IoBNT domain, the primary constraint has been manual or multi-stage feature engineering as is the case in [25][26][27][28][29][30][31], as well as limited dataset diversity as in [33]. In the IoT domain, utilize semi-hybrid classification strategies to detect network anomalies.…”

“…Extensive multi-stage feature engineering prior to DL application results in significant computational overhead which would be counter-intuitive for deployment in IDS, and IPS if re-training were required. Furthermore, supervised feature selection represented in prior schemes necessitates a new cycle of data collection, annotation, and training for updating classifiers [26][27][28][29][30][31]. However, such a training primitive is not applicable for the proposed hybrid since the addition of new data would only require preparation (pre-processing), with convolutional layers performing necessary feature selection.…”

Review of Physical Layer Security in Molecular Internet of Nano-Things

“…Unique secret key generation [25] × Vulnerable to Stronger Attacker communications. This requires insertion, localization [32], evading anomaly detection [33], and tampering to be jointly successful. There are a myriad of detailed scenarios on how molecularbased IoNT systems can be attacked and we detail their corresponding recent research in Table I.…”