Dependability in Pervasive Computing

Introduction

Pervasive Computing can be seen as an enabler technology for many new and exciting applications. Indicative examples of envisaged applications include the following: a smart shopping list, which automatically adds items with low supplies to the list; a digital personal assistant, which provides useful information on the basis of the user’s current context and needs (e.g. time schedules); a medical diagnostics system, which supervises vital functions of patients at all times and therefore allows much better diagnostics; a similar system could also be used to monitor heart patients and trigger emergency calls. Even closer in the future are systems, which allow continuous monitoring and tracking of goods. Just think of smart production, logistic or transportation systems. If for example luggage was to be equipped with RFID tags, long and time-consuming check-in procedures could be omitted, lost luggage would be reduced and customs would be simplified.

All these systems offer a lot of new interesting and useful possibilities, but would not be possible without Pervasive Computing technologies. However, as with every new system (generation) the question of dependability arises as soon as the idea of the system has sprung. Dependability can be typically divided into a number of aspects, namely (functional) correctness, safety, security, reliability, availability, transparency and traceability. It is interesting to note, that for most existing systems only a very limited number of aspects of dependability are relevant. For example a train control system must be able to tolerate component failures and fulfill its function at all times. Thus it must be safe, reliable and available. Aspects like security, transparency and traceability play little to no role. On the other hand, e-shopping systems need to be secure, transparent and traceable. Safety and reliability are not of much concern (as these systems have virtually no inherent potential for damage).

This is different for Pervasive Computing systems. Most Pervasive Computing systems come with requirements from all aspects of dependability. The reason for this lies in the nature of Pervasive Computing itself. Typically Pervasive Computing systems are very tightly connected with specific users (let it be directly such as a medical diagnostic system or indirectly such as a smart logistics system, which handles personal luggage at airports). Pervasive Computing systems often gather and store information on behavior, context, habits and plans of their users. This information forms the basis for many of the benefits that the system can offer to the users. For example, a heart attack warning system can only trigger emergency calls, if it knows the person it is monitoring and its position. Check-in procedures can only be circumvented, if pieces of luggage carry information about their owner respectively their target destination. Additionally, Pervasive Computing systems often rely on wireless communication. Considering these issues in parallel automatically raises requirements in terms of security, transparency and traceability. Secondly, Pervasive Computing systems are commonly embedded in the environment not only for gathering information, but also for making decisions or at least for decision support. Regarding safety, any decision can be either harmful or not. As a rule of thumb, the closer the connection to the physical world is, the higher the potential for a safety relevant impact will be. A lost emergency signal from a patient’s heart monitoring system might cost her live; wrong loading of transportation containers could make airplanes very instable and cause severe problems during flight. In addition, intended attacks and manipulations of the systems can cause severe consequences (let it be monetary or in the form of injuries or deaths). For example consider the case of smuggling dangerous luggage into airplanes by means of abusing the check-in system. Therefore, in general Pervasive Computing systems also need to be safe, reliable and functionally correct.