This chapter has discussed network security and robustness as a single, intricate problem. Although it may be argued that boosting the defensive capabilities of distributed systems and increasing their operational stability are different agendas in terms of the techniques involved, it is our opinion that they should not be separated. Indeed, ultimately, they both aim at the same target improving network survivability so as to ensure that acceptable quality of service can be maintained under stress, whether accidental or deliberate. As systems regrouping large numbers of interconnected devices become simultaneously more complex and dynamic, attack strategies based on exploiting architectural vulnerabilities in order to cause damage far beyond originally targeted nodes will become increasingly common. In a sense, the infamous distributed denial of service attack, whereby a pool of hijacked devices are enrolled in an attempt to flood a network with illegitimate service requests, can be regarded as an early example of this tactic being put into practice. We argue that the inextricable relationship between security and architecture has to be much more widely accepted on the defenders side if we are to devise suitable ways of protecting what is rapidly becoming a planetary, always-on information processing entity whose development (or blueprint) cannot be controlled. We think we have given enough evidence that adaptive perimeter techniques can successfully be combined with inventive design in order to reach a level of resilience to attack and possible side-effects far superior to that achievable by following either of these routes.
TABLE OF CONTENTS 
