Sergio Pozo

The management and automation of business processes have become an essential tasks within IT organizations. Companies could deploy business process management systems to automatize their business processes. BPMS needs to ensure that those are as dependable as possible. Fault tolerance techniques provide mechanisms to decrease the risk of possible faults in systems. In this paper, a framework for developing business processes with fault tolerance capabilities is provided. The framework presents different solutions in the fault tolerance scope. The solutions have been developed using a practical example and some results have been obtained, compared and discussed.

Developing and managing firewall Access Control Lists (ACLs) are hard, time-consuming, and error-prone tasks for a variety of reasons. Complexity of networks is constantly increasing, as it is the size of firewall ACLs. Networks have different access control requirements which must be translated by a network administrator into firewall ACLs. During this task, inconsistent rules can be introduced in the ACL. Furthermore, each time a rule is modified (e.g. updated, corrected when a fault is found, etc.) a new inconsistency with other rules can be introduced. An inconsistent firewall ACL implies, in general, a design or development fault, and indicates that the firewall is accepting traffic that should be denied or vice versa. In this paper we propose a complete and minimal consistency diagnosis process which has worst-case quadratic time complexity with the number of rules in a set of inconsistent rules. There are other proposals of consistency diagnosis algorithms. However they have different problems which can prevent their use with big, real-life, ACLs: on the one hand, the minimal ones have exponential worst-case time complexity; on the other hand, the polynomial ones are not minimal.

A Heuristic Process for Local Inconsistency Diagnosis in Firewall Rule Sets

The design and management of firewall ACLs is a very hard and error-prone task. Part of this complexity comes from the fact that each firewall platform has its own low-level language with a different functionality, syntax, and development environment. Although high-level languages have been proposed to model firewall ACLs, none of them has been widely adopted by the industry due to a combination of factors: high complexity, no support of important features of firewalls, etc. In this paper the most important access control policy languages are reviewed, with special focus on the development of firewall ACLs. Based on this analysis, a new domain specific language for firewall ACLs (AFPL2) is proposed, supporting more features that other languages do not cover (e.g. NAT). As the result of our design methodology, AFPL2 is very lightweight and easy to use. AFPL2 can be translated to existing low-level firewall languages, or be directly interpreted by firewall platforms, and is an extension to a previously developed language.

Firewalls provide the first line of defence of nearly all networked institutions today. However, Firewall ACLs could have inconsistencies, allowing traffic that should be denied or vice versa. In this paper, we analyze the inconsistency characterization problem as a separate problem of the diagnosis one, and propose formal definitions in order to characterize one-to- many inconsistencies. We identify the combinatorial part of the problem that generates exponential complexities in combined diagnosis and characterization algorithms proposed by other authors. Then we propose a decomposition of the combinatorial problem in several smaller combinatorial ones, which can effectively reduce the complexity of the problem. Finally, we propose an approximate heuristic and algorithms to solve the problem in worst case polynomial time. Although many algorithms have been proposed to address this problem, all of them are combinatorial. The presented algorithms are an heuristic way to solve the problem with polynomial complexity. There are no constraints on how rule field ranges are expressed.