Training general agents to follow complex instructions (tasks) in intricate environments (levels) remains a core challenge in reinforcement learning. Random sampling of task-level pairs often produces unsolvable combinations, highlighting the need to co-design tasks and levels. While unsupervised environment design (UED) has proven effective at automatically designing level curricula, prior work has only considered a fixed task. We present ATLAS (Aligning Tasks and Levels for Autocurricula of Specifications), a novel method that generates joint autocurricula over tasks and levels. Our approach builds upon UED to automatically produce solvable yet challenging task-level pairs for policy training. To evaluate ATLAS and drive progress in the field, we introduce an evaluation suite that models tasks as reward machines in Minigrid levels. Experiments demonstrate that ATLAS vastly outperforms random sampling approaches, particularly when sampling solvable pairs is unlikely. We further show that mutations leveraging the structure of both tasks and levels accelerate convergence to performant policies.

Warning: This paper contains content that may be inappropriate, offensive, or harmful.Automated red-teaming has emerged as a scalable approach for auditing Large Language Models (LLMs) prior to deployment, yet existing approaches lack mechanisms to efficiently adapt to model-specific vulnerabilities at inference.We introduce Red-Bandit, a red-teaming framework that adapts online to identify and exploit model failure modes under distinct attack styles (e.g., manipulation, slang).Red-Bandit post-trains a set of parameter-efficient LoRA experts, each specialized for a particular attack style, using reinforcement learning that rewards the generation of unsafe prompts via a rule-based safety model.At inference, a multiarmed bandit policy dynamically selects among these attack-style experts based on the target model's response safety, balancing exploration and exploitation.Red-Bandit outperforms stateof-the-art methods on AdvBench and Harm-Bench, achieving higher attack success rates under sufficient exploration budgets (ASR@10), while generating more human-readable adversarial prompts (lower perplexity).In addition, Red-Bandit's bandit policy serves as a diagnostic tool for identifying model-specific vulnerabilities by indicating which attack styles most effectively elicit harmful behaviors.

Reactive systems, which maintain ongoing interactions with their environment, are typically modeled using ω-regular languages. These languages characterize system behavior with infinite-length execution traces and are represented as nondeterministic Büchi automata (NBAs) or ω-regular expressions. Existing methods for synthesizing expressions from NBAs only handle state-based acceptance. This limitation forces the conversion of compact transition-based NBAs into larger state-based NBAs before synthesis. This article introduces the first direct synthesis method for ω-regular expressions from transition-based NBAs, eliminating the need for this transformation. The method works by decomposing an NBA into triplets of nondeterministic finite automata, subsuming existing pair-based decompositions to handle transition-based acceptance. We prove that our decomposition is correct, establishing our method’s soundness and completeness. We discuss the time and descriptional complexity of our method. Our empirical evaluation on 185 linear temporal logic formulas supports our hypothesis; transition-based synthesis reduces postfix size by 4.1× and 1.5× for recurrence and reactivity properties, respectively. We also analyze the structural NBA factors that determine when the transition-based NBA will yield a more compact expression and develop this into a simple criterion that, when applied, yields expressions that are at least as compact as those obtained by direct state-based synthesis.

Abstract Inferring linear temporal logic over finite traces ($\text{LTL}_{\text{f}}$) formulas from a set of example traces, known as passive learning, presents significant challenges due to its combinatorial nature. In this paper, we introduce a novel approach to $\text{LTL}_{\text{f}}$ passive learning based on inductive logic programming (ILP), leveraging the inductive learning of answer set programs framework. Our ILP-based method effectively exploits the set of example traces to guide the learning process, and experimental results demonstrate that it o ffers a more efficient solution compared to traditional techniques based on propositional satisfiability.

Neural Disjunctive Normal Form (DNF) based models are powerful and interpretable approaches to neuro-symbolic learning and have shown promising results in classification and reinforcement learning settings without prior knowledge of the tasks. However, their performance is degraded by the thresholding of the post-training symbolic translation process. We show here that part of the performance degradation during translation is due to its failure to disentangle the learned knowledge represented in the form of the networks’ weights. We address this issue by proposing a new disentanglement method; by splitting nodes that encode nested rules into smaller independent nodes, we are able to better preserve the models’ performance. Through experiments on binary, multiclass, and multilabel classification tasks (including those requiring predicate invention), we demonstrate that our disentanglement method provides compact and interpretable logical representations for the neural DNF-based models, with performance closer to that of their pre-translation counterparts. Our code is available at https://github.com/kittykg/disentangling-ndnf-classification.

Teams of embodied AI-enabled agents are critical for applications in extreme and highly dynamic environments. Developing robust controllers for such agents requires a deep understanding of the challenges encountered when attempting to coordinate and synchronize their individual perception-cognition-communication-action (PCCA) loops for team-wide mission objectives. We introduce a framework to explore the coordination of the PCCA loops across multiple agents in a new simulated physical environment designed to explore collaboration in each PCCA stage. This environment tasks teams of agents with the correct disposal of dangerous objects in an area and forces careful coordination of sensing, communication, movement, and manipulation actions by providing spatially-bounded communication, incorporating situations that require concerted effort by groups of agents, and introducing uncertainty into agents’ sensing capabilities. We provide a set of heuristic controllers, an offline oracle model, and an initial exploration of a Reward Machine-based controller that learns its policies from training. Together these approaches serve to provide insights into the complexity of the multi-agent PCCA loop coordination problem. The multiagent PCCA simulation environment, which supports AI and human-controlled agents, and the code for various agent controllers are available at https://github.com/nesl/AI-Collab.

This paper presents PROB-IRM, an approach that learns robust reward machines (RMs) for reinforcement learning (RL) agents from noisy execution traces. The key aspect of RM-driven RL is the exploitation of a finite-state ma- chine that decomposes the agent’s task into different sub- tasks. PROB-IRM uses a state-of-the-art inductive logic pro- gramming framework robust to noisy examples to learn RMs from noisy traces using the Bayesian posterior degree of be- liefs, thus ensuring robustness against inconsistencies. Piv- otal for the results is the interleaving between RM learning and policy learning: a new RM is learned whenever the RL agent generates a trace that is believed not to be accepted by the current RM. To speed up the training of the RL agent, PROB-IRM employs a probabilistic formulation of reward shaping that uses the posterior Bayesian beliefs derived from the traces. Our experimental analysis shows that PROB-IRM can learn (potentially imperfect) RMs from noisy traces and exploit them to train an RL agent to solve its tasks success- fully. Despite the complexity of learning the RM from noisy traces, agents trained with PROB-IRM perform comparably to agents provided with handcrafted RMs.

Technology advances in areas such as sensors, IoT, and robotics, enable new collaborative applications (e.g., autonomous devices). A primary requirement for such collaborations is to have a secure system that enables information sharing and information flow protection. A policy-based management system is a key mechanism for secure selective sharing of protected resources. However, policies in each party of a collaborative environment cannot be static as they have to adapt to different contexts and situations. One advantage of collaborative applications is that each party in the collaboration can take advantage of the knowledge of the other parties for learning or enhancing its own policies. We refer to this learning mechanism as policy transfer. The design of a policy transfer framework has challenges, including policy conflicts and privacy issues. Policy conflicts typically arise because of differences in the obligations of the parties, whereas privacy issues result because of data sharing constraints for sensitive data. Hence, the policy transfer framework should be able to tackle such challenges by considering minimal sharing of data and supporting policy adaptation to address conflict. In the paper, we propose a framework that aims at addressing such challenges. We introduce a formal definition of the policy transfer problem for attribute-based access control policies. We then introduce the transfer methodology which consists of three sequential steps. Finally, we report experimental results.

Large pre-trained language models such as BERT have been widely used as a framework for natural language understanding (NLU) tasks.However, recent findings have revealed that pre-trained language models are insensitive to word order.The performance on NLU tasks remains unchanged even after randomly permuting the word of a sentence, where crucial syntactic information is destroyed.To help preserve the importance of word order, we propose a simple approach called FORCED INVALIDA-TION (FI): forcing the model to identify permuted sequences as invalid samples.We perform an extensive evaluation of our approach on various English NLU and QA based tasks over BERT-based and attention-based models over word embeddings.Our experiments demonstrate that FI significantly improves the sensitivity of the models to word order. 1

Reward machines (RMs) are a recent formalism for representing the reward function of a reinforcement learning task through a finite-state machine whose edges encode subgoals of the task using high-level events. The structure of RMs enables the decomposition of a task into simpler and independently solvable subtasks that help tackle long-horizon and/or sparse reward tasks. We propose a formalism for further abstracting the subtask structure by endowing an RM with the ability to call other RMs, thus composing a hierarchy of RMs (HRM). We exploit HRMs by treating each call to an RM as an independently solvable subtask using the options framework, and describe a curriculum-based method to learn HRMs from traces observed by the agent. Our experiments reveal that exploiting a handcrafted HRM leads to faster convergence than with a flat HRM, and that learning an HRM is feasible in cases where its equivalent flat representation is not.

One of the ultimate goals of Artificial Intelligence is to assist humans in complex decision making. A promising direction for achieving this goal is Neuro-Symbolic AI, which aims to combine the interpretability of symbolic techniques with the ability of deep learning to learn from raw data. However, most current approaches require manually engineered symbolic knowledge, and where end-to-end training is considered, such approaches are either restricted to learning definite programs, or are restricted to training binary neural networks. In this paper, we introduce Neuro-Symbolic Inductive Learner (NSIL), an approach that trains a general neural network to extract latent concepts from raw data, whilst learning symbolic knowledge that maps latent concepts to target labels. The novelty of our approach is a method for biasing the learning of symbolic knowledge, based on the in-training performance of both neural and symbolic components. We evaluate NSIL on three problem domains of different complexity, including an NP-complete problem. Our results demonstrate that NSIL learns expressive knowledge, solves computationally complex problems, and achieves state-of-the-art performance in terms of accuracy and data efficiency. Code and technical appendix: https://github.com/DanCunnington/NSIL

Abstract Logic-based machine learning aims to learn general, interpretable knowledge in a data-efficient manner. However, labelled data must be specified in a structured logical form. To address this limitation, we propose a neural-symbolic learning framework, called Feed-Forward Neural-Symbolic Learner (FFNSL), that integrates a logic-based machine learning system capable of learning from noisy examples, with neural networks, in order to learn interpretable knowledge from labelled unstructured data. We demonstrate the generality of FFNSL on four neural-symbolic classification problems, where different pre-trained neural network models and logic-based machine learning systems are integrated to learn interpretable knowledge from sequences of images. We evaluate the robustness of our framework by using images subject to distributional shifts, for which the pre-trained neural networks may predict incorrectly and with high confidence. We analyse the impact that these shifts have on the accuracy of the learned knowledge and run-time performance, comparing FFNSL to tree-based and pure neural approaches. Our experimental results show that FFNSL outperforms the baselines by learning more accurate and interpretable knowledge with fewer examples.

This issue of TPLP contains selected papers of the 39 th

In the past decade, several systems for learning Answer Set Programs (ASP) have been proposed, including the recent FastLAS system. Compared to other state-of-the-art approaches to learning ASP, FastLAS is more scalable, as rather than computing the hypothesis space in full, it computes a much smaller subset relative to a given set of examples that is nonetheless guaranteed to contain an optimal solution to the task (called an OPT-sufficient subset). On the other hand, like many other Inductive Logic Programming (ILP) systems, FastLAS is designed to be run on a fixed learning task meaning that if new examples are discovered after learning, the whole process must be run again. In many real applications, data arrives in a stream. Rerunning an ILP system from scratch each time new examples arrive is inefficient. In this paper we address this problem by presenting IncrementalLAS, a system that uses a new technique, called hypothesis space expansion, to enable a FastLAS-like OPT-sufficient subset to be expanded each time new examples are discovered. We prove that this preserves FastLAS's guarantee of finding an optimal solution to the full task (including the new examples), while removing the need to repeat previous computations. Through our evaluation, we demonstrate that running IncrementalLAS on tasks updated with sequences of new examples is significantly faster than re-running FastLAS from scratch on each updated task.

We introduce Detect, Understand, Act (DUA), a neuro-symbolic reinforcement learning framework. The Detect component is composed of a traditional computer vision object detector and tracker. The Act component houses a set of options, high-level actions enacted by pre-trained deep reinforcement learning (DRL) policies. The Understand component provides a novel answer set programming (ASP) paradigm for effectively learning symbolic meta-policies over options using inductive logic programming (ILP). We evaluate our framework on the Animal-AI (AAI) competition testbed, a set of physical cognitive reasoning problems. Given a set of pre-trained DRL policies, DUA requires only a few examples to learn a meta-policy that allows it to improve the state-of-the-art on multiple of the most challenging categories from the testbed. DUA constitutes the first holistic hybrid integration of computer vision, ILP and DRL applied to an AAI-like environment and sets the foundations for further use of ILP in complex DRL challenges.

In the study of reasoning in neural networks, recent efforts have sought to improve consistency and coherence of sequence models, leading to important developments in the area of neuro-symbolic AI. In symbolic AI, the concepts of consistency and coherence can be defined and verified formally, but for neural networks these definitions are lacking. The provision of such formal definitions is crucial to offer a common basis for the quantitative evaluation and systematic comparison of connectionist, neuro-symbolic and transfer learning approaches. In this paper, we introduce formal definitions of consistency and coherence for neural systems. To illustrate the usefulness of our definitions, we propose a new dynamic relation-decoder model built around the principles of consistency and coherence. We compare our results with several existing relation-decoders using a partial transfer learning task based on a novel data set introduced in this paper. Our experiments show that relation-decoders that maintain consistency over unobserved regions of representation space retaincoherence across domains, whilst achieving better transfer learning performance.

Abstract In this paper we introduce Detect, Understand, Act (DUA), a neuro-symbolic reinforcement learning framework. The Detect component is composed of a traditional computer vision object detector and tracker. The Act component houses a set of options, high-level actions enacted by pre-trained deep reinforcement learning (DRL) policies. The Understand component provides a novel answer set programming (ASP) paradigm for symbolically implementing a meta-policy over options and effectively learning it using inductive logic programming (ILP). We evaluate our framework on the Animal-AI (AAI) competition testbed, a set of physical cognitive reasoning problems. Given a set of pre-trained DRL policies, DUA requires only a few examples to learn a meta-policy that allows it to improve the state-of-the-art on multiple of the most challenging categories from the testbed. DUA constitutes the first holistic hybrid integration of computer vision, ILP and DRL applied to an AAI-like environment and sets the foundations for further use of ILP in complex DRL challenges.

Numerical reasoning based machine reading comprehension is a task that involves reading comprehension along with using arithmetic operations such as addition, subtraction, sorting, and counting.The DROP benchmark (Dua et al., 2019) is a recent dataset that has inspired the design of NLP models aimed at solving this task.The current standings of these models in the DROP leaderboard, over standard metrics, suggest that the models have achieved near-human performance.However, does this mean that these models have learned to reason?In this paper, we present a controlled study on some of the top-performing model architectures for the task of numerical reasoning.Our observations suggest that the standard metrics are incapable of measuring progress towards such tasks.

Recently, novel ILP systems under the answer set semantics have been proposed, some of which are robust to noise and scalable over large hypothesis spaces. One such system is FastLAS, which is significantly faster than other state-of-the-art ASP-based ILP systems. FastLAS is, however, only capable of Observational Predicate Learning (OPL), where the learned hypothesis defines predicates that are directly observed in the examples. It cannot learn knowledge that is indirectly observable, such as learning causes of observed events. This class of problems, known as non-OPL, is known to be difficult to handle in the context of non-monotonic semantics. Solving non-OPL learning tasks whilst preserving scalability is a challenging open problem. We address this problem with a new abductive method for translating examples of a non-OPL task to a set of examples, called possibilities, such that the original example is covered iff at least one of the possibilities is covered. This new method allows an ILP system capable of performing OPL tasks to be "upgraded" to solve non-OPL tasks. In particular, we present our new FastNonOPL system, which upgrades FastLAS with the new possibility generation. We compare it to other state-of-the-art ASP-based ILP systems capable of solving non-OPL tasks, showing that FastNonOPL is significantly faster, and in many cases more accurate, than these other systems.

In this paper we present ISA, an approach for learning and exploiting subgoals in episodic reinforcement learning (RL) tasks. ISA interleaves reinforcement learning with the induction of a subgoal automaton, an automaton whose edges are labeled by the task’s subgoals expressed as propositional logic formulas over a set of high-level events. A subgoal automaton also consists of two special states: a state indicating the successful completion of the task, and a state indicating that the task has finished without succeeding. A state-of-the-art inductive logic programming system is used to learn a subgoal automaton that covers the traces of high-level events observed by the RL agent. When the currently exploited automaton does not correctly recognize a trace, the automaton learner induces a new automaton that covers that trace. The interleaving process guarantees the induction of automata with the minimum number of states, and applies a symmetry breaking mechanism to shrink the search space whilst remaining complete. We evaluate ISA in several gridworld and continuous state space problems using different RL algorithms that leverage the automaton structures. We provide an in-depth empirical analysis of the automaton learning performance in terms of the traces, the symmetry breaking and specific restrictions imposed on the final learnable automaton. For each class of RL problem, we show that the learned automata can be successfully exploited to learn policies that reach the goal, achieving an average reward comparable to the case where automata are not learned but handcrafted and given beforehand.

In this work we present ISA, a novel approach for learning and exploiting subgoals in reinforcement learning (RL). Our method relies on inducing an automaton whose transitions are subgoals expressed as propositional formulas over a set of observable events. A state-of-the-art inductive logic programming system is used to learn the automaton from observation traces perceived by the RL agent. The reinforcement learning and automaton learning processes are interleaved: a new refined automaton is learned whenever the RL agent generates a trace not recognized by the current automaton. We evaluate ISA in several gridworld problems and show that it performs similarly to a method for which automata are given in advance. We also show that the learned automata can be exploited to speed up convergence through reward shaping and transfer learning across multiple tasks. Finally, we analyze the running time and the number of traces that ISA needs to learn an automata, and the impact that the number of observable events have on the learner's performance.

Inductive Logic Programming (ILP) systems aim to find a set of logical rules, called a hypothesis, that explain a set of examples. In cases where many such hypotheses exist, ILP systems often bias towards shorter solutions, leading to highly general rules being learned. In some application domains like security and access control policies, this bias may not be desirable, as when data is sparse more specific rules that guarantee tighter security should be preferred. This paper presents a new general notion of a scoring function over hypotheses that allows a user to express domain-specific optimisation criteria. This is incorporated into a new ILP system, called FastLAS, that takes as input a learning task and a customised scoring function, and computes an optimal solution with respect to the given scoring function. We evaluate the accuracy of FastLAS over real-world datasets for access control policies and show that varying the scoring function allows a user to target domain-specific performance metrics. We also compare FastLAS to state-of-the-art ILP systems, using the standard ILP bias for shorter solutions, and demonstrate that FastLAS is significantly faster and more scalable.

We introduce a novel approach for the computation of stable and supported models of normal logic programs in continuous vector spaces by a gradient-based search method. Specifically, the application of the immediate consequence operator of a program reduct can be computed in a vector space. To do this, Herbrand interpretations of a propositional program are embedded as 0-1 vectors in $\mathbb{R}^N$ and program reducts are represented as matrices in $\mathbb{R}^{N \times N}$. Using these representations we prove that the underlying semantics of a normal logic program is captured through matrix multiplication and a differentiable operation. As supported and stable models of a normal logic program can now be seen as fixed points in a continuous space, non-monotonic deduction can be performed using an optimisation process such as Newton's method. We report the results of several experiments using synthetically generated programs that demonstrate the feasibility of the approach and highlight how different parameter values can affect the behaviour of the system.

Human reasoning involves recognising common underlying principles across many examples. The by-products of such reasoning are invariants that capture patterns such as "if someone went somewhere then they are there", expressed using variables "someone" and "somewhere" instead of mentioning specific people or places. Humans learn what variables are and how to use them at a young age. This paper explores whether machines can also learn and use variables solely from examples without requiring human pre-engineering. We propose Unification Networks, an end-to-end differentiable neural network approach capable of lifting examples into invariants and using those invariants to solve a given task. The core characteristic of our architecture is soft unification between examples that enables the network to generalise parts of the input into variables, thereby learning invariants. We evaluate our approach on five datasets to demonstrate that learning invariants captures patterns in the data and can improve performance over baselines.

There has been over the past decade a rapid change towards computational environments that are comprised of large and diverse sets of devices, many of them mobile, which can connect in flexible and context-dependent ways. Examples range from networks where we can have communications between powerful cloud centers, to the myriad of simple sensor devices on the IoT. As the management of these dynamic environments becomes ever more complex, we want to propose policy migrations as a methodology to simplify the management of security policies by re-utilizing and re-deploying existing policies as the systems change. We are interested in understanding the challenges raised answering the following question: given a security policy that is being enforced in a particular source computational device, what does it entail to migrate this policy to be enforced in a different target device? Because of the differences between devices and because these devices cannot be seen in isolation but in the context where they are deployed, the meaning of the policy enforced in the source device needs to be re-interpreted and implemented in the context of the target device. The aim of the paper is to present a formal framework to evaluate the appropriateness of the migration.

Three kinds of submissions were accepted:• Technical papers for technically sound, innovative ideas that can advance the state of logic programming.• Application papers that impact interesting application domains.• System and tool papers which emphasize novelty, practicality, usability, and availability of the systems and tools described.

In this paper we introduce an extension of context-free grammars called answer set grammars (ASGs). These grammars allow annotations on production rules, written in the language of Answer Set Programming (ASP), which can express context-sensitive constraints. We investigate the complexity of various classes of ASG with respect to two decision problems: deciding whether a given string belongs to the language of an ASG and deciding whether the language of an ASG is non-empty. Specifically, we show that the complexity of these decision problems can be lowered by restricting the subset of the ASP language used in the annotations. To aid the applicability of these grammars to computational problems that require context-sensitive parsers for partially known languages, we propose a learning task for inducing the annotations of an ASG. We characterise the complexity of this task and present an algorithm for solving it. An evaluation of a (prototype) implementation is also discussed.

To facilitate information sharing between systems and devices in a distributed environment, unstructured data from various sensors must be analysed accordingly. Recent work has developed the notion of a context-dependant generative policy framework capable of learning generative policy models from strings and text-based data in a tabular format. However, it is vital that unstructured contextual information can be analysed alongside tabular data, potentially at the edge of the network to enable generative policy models to be applied to more complex tasks. This paper performs a deep-dive into the field of neuralsymbolic machine learning with a view towards enabling future neural-symbolic generative policy models that are capable of analysing both structured and unstructured data, whilst providing full transparency and enabling edge of network reasoning capability. Firstly, an existing technique called DeepProbLog is investigated and applied to a policy inferencing task based on unstructured data and secondly, a recent inductive logic programming technique currently used for learning generative policy models is evaluated with respect to a policy learning task also based on unstructured data. Finally, the results of both tasks are discussed to provide a platform for future research into enabling neural-symbolic generative policy models.

When creating real-world machine learning applications, system developers have to deal with the challenges of a dynamic environment where conditions change frequently, data has uncertainty, and new unanticipated situations are encountered. This requires a flexible approach in deciding how to use, adapt and create an AI model. Ensemble learning, where multiple models are trained, and use concurrently provides one way to address some of the issues. However, ensembles as used within the AI literature have primarily focused on creating a better model in a static environment. If we couple ensemble models with the concept of policy based control, we can create a system that is able to deal better with real-world scenarios. If we further augment the system so that it can generate its own policies, we can make progress towards the goal of a broad AI which can dynamically adapt itself. In this paper, we present an architecture for policy based ensemble, and show how it can lead to an approach towards broad AI.

Policy systems are critical for managing missions and collaborative activities carried out by coalitions involving different organizations. Conventional policy-based management approaches are not suitable for next-generation coalitions that will involve not only humans, but also autonomous computing devices and systems. It is critical that those parties be able to generate and customize policies based on contexts and activities. This paper introduces a novel approach for the autonomic generation of policies by autonomous parties. The framework combines context free grammars, answer set programs, and inductionbased learning. It allows a party to generate its own policies, based on a grammar and some semantic constraints, by learning from examples. The paper also outlines initial experiments in the use of such a symbolic approach and outlines relevant research challenges, ranging from explainability to quality assessment of policies.

Generative Policy Models (GPMs) have been proposed as a method for future autonomous decision making in a distributed, collaborative environment. To learn a GPM, previous policy examples that contain policy features and the corresponding policy decisions are used. Recently, GPMs have been constructed using both symbolic and statistical learning algorithms. In either case, the goal of the learning process is to create a model across a wide range of contexts from which specific policies may be generated in a given context. Empirically, we expect each learning approach to provide certain advantages over the other. This paper assesses the relative performance of each learning approach in order to examine these advantages and disadvantages. Several carefully prepared data sets are used to train a variety of models across different learning algorithms, where models for each learning algorithm are trained with varying amounts of labelled examples. The performance of each model is evaluated across a variety of metrics which indicates the strength of each learning algorithm for the different scenarios presented and the amount of training data provided. Finally, future research directions are outlined to fully realise GPMs in a distributed, collaborative environment.

Generative policies have recently been researched to provide a method for next generation security policies. They are created using either traditional machine learning techniques or, more recently, inductive learning of answer set programs. The latter method is targeted to the learning of Answer Set Grammars (ASG), a new notion of generative policy model for security policies that has the benefit of transparent explainability of the learned outcomes. This paper proposes a military scenario based on logistical resupply from a military base to coalition forces located in a nearby urban area or city. We describe the scenario and accompanying policy such that the context of the resupply missions (and therefore the policy) changes over time. The set of policies and related changes over time have been manually defined using a set of human created rules to replicate how security policies would currently be created by humans in such scenarios. We show how inductive learning of answer set programs can successfully learn ASG generative policy models that capture the human-driven rules from just example traces and decisions made at different time points and with respect to different contextual situations that can arise during the resupply mission. These results demonstrate the utility of ASG generative policy as a method for modelling human-driven policy rules.

Traditionally most of the work in the field of Inductive Logic Programming (ILP) has addressed the problem of learning Prolog programs. On the other hand, Answer Set Programming is increasingly being used as a powerful language for knowledge representation and reasoning, and is also gaining increasing attention in industry. Consequently, the research activity in ILP has widened to the area of Answer Set Programming, witnessing the proposal of several new learning frameworks that have extended ILP to learning answer set programs. In this paper, we investigate the theoretical properties of these existing frameworks for learning programs under the answer set semantics. Specifically, we present a detailed analysis of the computational complexity of each of these frameworks with respect to the two decision problems of deciding whether a hypothesis is a solution of a learning task and deciding whether a learning task has any solutions. We introduce a new notion of generality of a learning framework, which enables us to define a framework to be more general than another in terms of being able to distinguish one ASP hypothesis solution from a set of incorrect ASP programs. Based on this notion, we formally prove a generality relation over the set of existing frameworks for learning programs under answer set semantics. In particular, we show that our recently proposed framework, Context-dependent Learning from Ordered Answer Sets, is more general than brave induction, induction of stable models, and cautious induction, and maintains the same complexity as cautious induction, which has the highest complexity of these frameworks.

With the advancement in the technology, deploying connected assets - especially intelligent autonomous assets - to obtain the evolving picture of dynamic environments are fast becoming a reality - and a need - for effective and efficient decision making. In such environments, these assets need to function in unison with each other to achieve the goals, and especially in a collaborative environments (e.g., coalition environments) they need to respect the constraints placed on them by the collective as well as by the owner parties. Typically, policies are used to govern such constraints and interactions, but the existing state-of-the-art relies on predefined user policies to achieve the effect, which is not scalable nor practical in collaborative and dynamic environments. Motivated by this observation and the recent uptake in learning technologies, in this paper, we present our vision on a framework that can (a) employ multiple techniques to create domain knowledge that can help assets to determine which policies are critical for which context, how to solve conflicts among policies, and how to autonomously generate and refine existing policies; (b) represent knowledge in a localized wiki-like approach so that fault tolerant knowledge discovery is supported; (c) provide efficient query interface for assets to discover needed knowledge in a secure manner; and (d) contextualize knowledge so as to enable other similar assets to quickly bootstrap or initialize themselves in unknown contexts when new events occur.

Some online social networks (OSNs) allow users to define friendship-groups as reusable shortcuts for sharing information with multiple contacts. Posting exclusively to a friendship-group gives some privacy control, while supporting communication with (and within) this group. However, recipients of such posts may want to reuse content for their own social advantage, and can bypass existing controls by copy-pasting into a new post; this cross-posting poses privacy risks. This paper presents a learning to share approach that enables the incorporation of more nuanced privacy controls into OSNs. Specifically, we propose a reusable, adaptive software architecture that uses rigorous runtime analysis to help OSN users to make informed decisions about suitable audiences for their posts. This is achieved by supporting dynamic formation of recipient-groups that benefit social interactions while reducing privacy risks. We exemplify the use of our approach in the context of Facebook.

With the introduction of network function virtualization technology, migrating entire enterprise data centers into the cloud has become a possibility. However, for a cloud service provider (CSP) to offer such services, several research problems still need to be addressed. In previous work, we have introduced a platform, called network function center (NFC), to study research issues related to virtualized network functions (VNFs). In an NFC, we assume VNFs to be implemented on virtual machines that can be deployed in any server in the CSP network. We have proposed a resource allocation algorithm for VNFs based on genetic algorithms (GAs). In this paper, we present a comprehensive analysis of two resource allocation algorithms based on GA for: 1) the initial placement of VNFs and 2) the scaling of VNFs to support traffic changes. We compare the performance of the proposed algorithms with a traditional integer linear programming resource allocation technique. We then combine data from previous empirical analyses to generate realistic VNF chains and traffic patterns, and evaluate the resource allocation decision making algorithms. We assume different architectures for the data center, implement different fitness functions with GA, and compare their performance when scaling over the time.

peer reviewed

Requirements incompleteness is often the result of unanticipated adverse conditions which prevent the software and its environment from behaving as expected. These conditions represent risks that can cause severe software failures. The identification and resolution of such risks is therefore a crucial step towards requirements completeness. Obstacle analysis is a goal-driven form of risk analysis that aims at detecting missing conditions that can obstruct goals from being satisfied in a given domain, and resolving them.

Probabilistic programming is an area of research that aims to develop general inference algorithms for probabilistic models expressed as probabilistic programs whose execution corresponds to inferring the parameters of those models. In this paper, we introduce a probabilistic programming language (PPL) based on abductive logic programming for performing inference in probabilistic models involving categorical distributions with Dirichlet priors. We encode these models as abductive logic programs enriched with probabilistic definitions and queries, and show how to execute and compile them to boolean formulas. Using the latter, we perform generalized inference using one of two proposed Markov Chain Monte Carlo (MCMC) sampling algorithms: an adaptation of uncollapsed Gibbs sampling from related work and a novel collapsed Gibbs sampling (CGS). We show that CGS converges faster than the uncollapsed version on a latent Dirichlet allocation (LDA) task using synthetic data. On similar data, we compare our PPL with LDA-specific algorithms and other PPLs. We find that all methods, except one, perform similarly and that the more expressive the PPL, the slower it is. We illustrate applications of our PPL on real data in two variants of LDA models (Seed and Cluster LDA), and in the repeated insertion model (RIM). In the latter, our PPL yields similar conclusions to inference with EM for Mallows models.

Researchers have recently shown that declarative database query languages, such as Datalog, could naturally be used to specify and implement network protocols and services. In this paper, we present a declarative framework for the specification, execution, simulation, and analysis of distributed applications. Distributed applications, including routing protocols, can be specified using a Declarative Networking language, called D2C, whose semantics capture the notion of a Distributed State Machine (DSM), i.e., a network of computational nodes that communicate with each other through the exchange of data. The D2C specification can be directly executed using the DSM computational infrastructure of our framework. The same specification can be simulated and formally verified. The simulation component integrates the DSM tool within a network simulation environment and allows developers to simulate network dynamics and collect data about the execution in order to evaluate application responses to network changes. The formal analysis component of our framework, instead, complements the empirical testing by supporting the verification of different classes of properties of distributed algorithms, including convergence of network routing protocols. To demonstrate the generality of our framework, we show how it can be used to analyze two classes of network routing protocols, a path vector and a Mobile Ad-Hoc Network (MANET) routing protocol, and execute a distributed algorithm for pattern formation in multi-robot systems.

Abstract In recent years, several frameworks and systems have been proposed that extend Inductive Logic Programming (ILP) to the Answer Set Programming (ASP) paradigm. In ILP, examples must all be explained by a hypothesis together with a given background knowledge. In existing systems, the background knowledge is the same for all examples; however, examples may be context-dependent. This means that some examples should be explained in the context of some information, whereas others should be explained in different contexts. In this paper, we capture this notion and present a context-dependent extension of theLearning from Ordered Answer Setsframework. In this extension, contexts can be used to further structure the background knowledge. We then propose a new iterative algorithm, ILASP2i, which exploits this feature to scale up the existing ILASP2 system to learning tasks with large numbers of examples. We demonstrate the gain in scalability by applying both algorithms to various learning tasks. Our results show that, compared to ILASP2, the newly proposed ILASP2i system can be two orders of magnitude faster and use two orders of magnitude less memory, whilst preserving the same average accuracy.

By allowing network functions to be virtualized and run on commodity hardware, NFV enables new properties (e.g., elastic scaling), and new service models for Service Providers, Enterprises, and Telecommunication Service Providers. However, for NFV to be offered as a service, several research problems still need to be addressed. In this paper, we focus and propose a new service chaining algorithm. Existing solutions suffer two main limitations: First, existing proposals often rely on mixed Integer Linear Programming to optimize VM allocation and network management, but our experiments show that such approach is too slow taking hours to find a solution. Second, although existing proposals have considered the VM placement and network configuration jointly, they frequently assume the network configuration cannot be changed. Instead, we believe that both computing and network resources should be able to be updated concurrently for increased flexibility and to satisfy SLA and Qos requirements. As such, we formulate and propose a Genetic Algorithm based approach to solve the VM allocation and network management problem. We built an experimental NFV platform, and run a set of experiments. The results show that our proposed GA approach can compute configurations to to three orders of magnitude faster than traditional solutions.

Signature-based network intrusion detection systems (S-IDS) have become an important security tool in the protection of an organisation's infrastructure against external intruders. By analysing network traffic, S-IDS' detect network intrusions. An organisation may deploy one or multiple S-IDS', each working independently with the assumption that it can monitor all packets of a given flow to detect intrusion signatures. However, emerging technologies (e.g., Multi-Path TCP) violate this assumption, as traffic can be concurrently sent across different paths (e.g., WiFi, Cellular) to boost network performance. Attackers may exploit this capability and split malicious payloads across multiple paths to evade traditional signature-based network intrusion detection systems. Although multiple monitors may be deployed, none of them has the full coverage of the network traffic to detect the intrusion signature. In this paper, we formalise this distributed signature-based intrusion detection problem as an asynchronous online exact string matching problem, and propose an algorithm for it. To demonstrate its effectiveness we conducted comprehensive experiments. Our results show that the behaviour of our algorithm depends only on the packet arrival rate: delay in detecting the signature grows linearly with respect to the packet arrival rate and with small communication overhead.

Abstract This paper contributes to the area of inductive logic programming by presenting a new learning framework that allows the learning of weak constraints in Answer Set Programming (ASP). The framework, calledLearning from Ordered Answer Sets, generalises our previous work on learning ASP programs without weak constraints, by considering a new notion of examples asorderedpairs of partial answer sets that exemplify which answer sets of a learned hypothesis (together with a given background knowledge) arepreferredto others. In this new learning task inductive solutions are searched within a hypothesis space of normal rules, choice rules, and hard and weak constraints. We propose a new algorithm, ILASP2, which is sound and complete with respect to our new learning framework. We investigate its applicability to learning preferences in an interview scheduling problem and also demonstrate that when restricted to the task of learning ASP programs without weak constraints, ILASP2 can be much more efficient than our previously proposed system.

Short term studies in controlled environments have shown that user behaviour is consistent enough to predict disruptive smartphone notifications. However, in practice, user behaviour changes over time (concept drift) and individual user preferences need to be considered. There is a lack of research on which methods are best suited for predicting disruptive smartphone notifications longer-term, taking into account varying error costs. In this paper we report on a 16 week field study comparing how well different learners perform at mitigating disruptive incoming phone calls.

The purpose of requirements validation is to determine whether a large requirements set will lead to the achievement of system-related goals under different conditions — a task that needs automation if it is to be performed quickly and accurately. One reason for the current lack of software tools to undertake such validation is the absence of the computational mechanisms needed to associate scenario, system specification and goal analysis tools. Therefore, in this paper, we report first research experiments in developing these new capabilities, and demonstrate them with a non-trivial example associated with a Rolls Royce aircraft engine software component.

Environment domain models are a key part of the information used by adaptive systems to determine their behaviour. These models can be incomplete or inaccurate. In addition, since adaptive systems generally operate in environments which are subject to change, these models are often also out of date. To update and correct these models, the system should observe how the environment responds to its actions, and compare these responses to those predicted by the model. In this paper, we use a probabilistic rule learning approach, NoMPRoL, to update models using feedback from the running system in the form of execution traces. NoMPRoL is a technique for nonmonotonic probabilistic rule learning based on a transformation of an inductive logic programming task into an equivalent abductive one. In essence, it exploits consistent observations by finding general rules which explain observations in terms of the conditions under which they occur. The updated models are then used to generate new behaviour with a greater chance of success in the actual environment encountered.

Abstract There is an increasing interest in using logic programming to specify and implement distributed algorithms, including a variety of network applications. These are applications where data and computation are distributed among several devices and where, in principle, all the devices can exchange data and share the computational results of the group. In this paper we propose a declarative approach to distributed computing whereby distributed algorithms and communication models can be (i) specified as action theories of fluents and actions; (ii) executed as collections of distributed state machines, where devices are abstracted as (input/output) automata that can exchange messages; and (iii) analysed using existing results on connecting causal theories and Answer Set Programming. Results on the application of our approach to different classes of network protocols are also presented.

The process of Requirements Engineering (RE) includes many activities, from goal elicitation to requirements specification. The aim is to develop an operational requirements specification that is guaranteed to satisfy the goals. In this paper, we propose a formal, systematic approach for generating a set of operational requirements that are complete with respect to given goals. We show how the integration of model checking and inductive learning can be effectively used to do this. The model checking formally verifies the satisfaction of the goals and produces counterexamples when incompleteness in the operational requirements is detected. The inductive learning process then computes operational requirements from the counterexamples and user-provided positive examples. These learned operational requirements are guaranteed to eliminate the counterexamples and be consistent with the goals. This process is performed iteratively until no goal violation is detected. The proposed framework is a rigorous, tool-supported requirements elaboration technique which is formally guided by the engineer's knowledge of the domain and the envisioned system.

This paper addresses the making of security decisions, such as access-control decisions or spam filtering decisions, under uncertainty, when the benefit of doing so outweighs the need to absolutely guarantee these decisions are correct. For instance, when there are limited, costly, or failed communication channels to a policy-decision-point. Previously, local caching of decisions has been proposed, but when a correct decision is not available, either a policy-decision-point must be contacted, or a default decision used. We improve upon this model by using learned classifiers of access control decisions. These classifiers, trained on known decisions, infer decisions when an exact match has not been cached, and uses intuitive notions of utility, damage and uncertainty to determine when an inferred decision is preferred over contacting a remote PDP. Clearly there is uncertainty in the predicted decisions, introducing a degree of risk. Our solution proposes a mechanism to quantify the uncertainty of these decisions and allows administrators to bound the overall risk posture of the system. The learning component continuously refines its models based on inputs from a central policy server in cases where the risk is too high or there is too much uncertainty. We have validated our models by building a prototype system and evaluating it with requests from real access control policies. Our experiments show that over a range of system parameters, it is feasible to use machine learning methods to infer access control policies decisions. Thus our system yields several benefits, including reduced calls to the PDP, reducing latency and communication costs; increased net utility; and increased system survivability.

An understanding of information flow has many applications, including for maximizing marketing impact on social media, limiting malware propagation, and managing undesired disclosure of sensitive information. This paper presents scalable methods for both learning models of information flow in networks from data, based on the Independent Cascade Model, and predicting probabilities of unseen flow from these models. Our approach is based on a principled probabilistic construction and results compare favourably with existing methods in terms of accuracy of prediction and scalable evaluation, with the addition that we are able to evaluate a broader range of queries than previously shown, including probability of joint and/or conditional flow, as well as reflecting model uncertainty. Exact evaluation of flow probabilities is exponential in the number of edges and naive sampling can also be expensive, so we propose sampling in an efficient Markov-Chain Monte-Carlo fashion using the Metropolis-Hastings algorithm -- details described in the paper. We identify two types of data, those where the paths of past flows are known -- attributed data, and those where only the endpoints are known -- unattributed data. Both data types are addressed in this paper, including training methods, example real world data sets, and experimental evaluation. In particular, we investigate flow data from the Twitter microblogging service, exploring the flow of messages through retweets (tweet forwards) for the attributed case, and the propagation of hash tags (metadata tags) and urls for the unattributed case.

Missing requirements are known to be among the major causes of software failure. They often result from a natural inclination to conceive over-ideal systems where the software-to-be and its environment always behave as expected. Obstacle analysis is a goal-anchored form of risk analysis whereby exceptional conditions that may obstruct system goals are identified, assessed and resolved to produce complete requirements. Various techniques have been proposed for identifying obstacle conditions systematically. Among these, the formal ones have limited applicability or are costly to automate. This paper describes a tool-supported technique for generating a set of obstacle conditions guaranteed to be complete and consistent with respect to the known domain properties. The approach relies on a novel combination of model checking and learning technologies. Obstacles are iteratively learned from counterexample and witness traces produced by model checking against a goal and converted into positive and negative examples, respectively. A comparative evaluation is provided with respect to published results on the manual derivation of obstacles in a real safety-critical system for which failures have been reported.

Over the last 15 years, many policy languages have been developed for specifying policies and credentials under the trust management paradigm. What has been missing is a formal semantics - in particular, one that would capture the inherently dynamic nature of trust management, where access decisions are based on the local policy in conjunction with varying sets of dynamically submitted credentials. The goal of this paper is to rest trust management on a solid formal foundation. To this end, we present a model theory that is based on Kripke structures for counterfactual logic. The semantics enjoys compositionality and full abstraction with respect to a natural notion of observational equivalence between trust management policies. Furthermore, we present a corresponding Hilbert-style axiomatization that is expressive enough for reasoning about a system's observables on the object level. We describe an implementation of a mechanization of the proof theory, which can be used to prove non-trivial meta-theorems about trust management systems, as well as analyze probing attacks on such systems. Our benchmark results show that this logic-based approach performs significantly better than the only previously available, ad-hoc analysis method for probing attacks.

Abstract In this paper we propose a use-case-driven iterative design methodology for normative frameworks, also called virtual institutions, which are used to govern open systems. Our computational model represents the normative framework as a logic program under answer set semantics (ASP). By means of an inductive logic programming approach, implemented using ASP, it is possible to synthesise new rules and revise the existing ones. The learning mechanism is guided by the designer who describes the desired properties of the framework through use cases, comprising (i) event traces that capture possible scenarios, and (ii) a state that describes the desired outcome. The learning process then proposes additional rules, or changes to current rules, to satisfy the constraints expressed in the use cases. Thus, the contribution of this paper is a process for the elaboration and revision of a normative framework by means of a semi-automatic and iterative process driven from specifications of (un)desirable behaviour. The process integrates a novel and general methodology for theory revision based on ASP.

The automation of policy refinement, whilst promising great benefits for policy-based management, has hitherto received relatively little treatment in the literature, with few concrete approaches emerging. In this paper we present initial steps towards a framework for automated distributed policy refinement for both obligation and authorization policies. We present examples drawn from military scenarios, describe details of our formalism and methods for action decomposition, and discuss directions for future research.

Abstract One of the difficulties in goal-oriented requirements engineering (GORE) is the construction of behaviour models from declarative goal specifications. This paper addresses this problem using a combination of model checking and machine learning. First, a goal model is transformed into a (potentially Zeno) behaviour model. Then, via an iterative process, Zeno traces are identified by model checking the behaviour model against a time progress property, and inductive logic programming (ILP) is used to learn operational requirements ( pre-conditions ) that eliminate these traces. The process terminates giving a non-Zeno behaviour model produced from the learned pre-conditions and the given goal model.

Despite several research studies, the effective analysis of policy based systems remains a significant challenge. Policy analysis should at least (i) be expressive (ii) take account of obligations and authorizations, (iii) include a dynamic system model, and (iv) give useful diagnostic information. We present a logic-based policy analysis framework which satisfies these requirements, showing how many significant policy-related properties can be analysed, and we give details of a prototype implementation.

Goal-oriented methods have increasingly been recognised as an effective means for eliciting, elaborating, analysing and specifying software requirements. A key activity in these approaches is the elaboration of a correct and complete set of opertional requirements, in the form of pre- and trigger-conditions, that guarantee the system goals. Few existing approaches provide support for this crucial task and mainly rely on significant effort and expertise of the engineer. In this paper we propose a tool-based framework that combines model checking, inductive learning and scenarios for elaborating operational requirements from goal models. This is an iterative process that requires the engineer to identify positive and negative scenarios from counterexamples to the goals, generated using model checking, and to select operational requirements from suggestions computed by inductive learning.

Firewalls remain the main perimeter security protection for corporate networks. However, network size and complexity make firewall configuration and maintenance notoriously difficult. Tools are needed to analyse firewall configurations for errors, to verify that they correctly implement security requirements and to generate configurations from higher-level requirements. In this paper we extend our previous work on the use of formal argumentation and preference reasoning for firewall policy analysis and develop means to automatically generate firewall policies from higher-level requirements. This permits both analysis and generation to be done within the same framework, thus accommodating a wide variety of scenarios for authoring and maintaining firewall configurations. We validate our approach by applying it to both examples from the literature and real firewall configurations of moderate size (ap 150 rules).

Policy-based management provides the ability to (re-)configure differentiated services networks so that desired Quality of Service (QoS) goals are achieved. This requires implementing network provisioning decisions, performing admission control, and adapting bandwidth allocation to emerging traffic demands. A policy-based approach facilitates flexibility and adaptability as policies can be dynamically changed without modifying the underlying implementation. However, inconsistencies may arise in the policy specification. In this paper we provide a comprehensive set of QoS policies for managing Differentiated Services (DiffServ) networks, and classify the possible conflicts that can arise between them. We demonstrate the use of Event Calculus and formal reasoning for the analysis of both static and dynamic conflicts in a semi-automated fashion. In addition, we present a conflict analysis tool that provides network administrators with a user-friendly environment for determining and resolving potential inconsistencies. The tool has been extensively tested with large numbers of policies over a range of conflict types.

E-learning is a general term used to refer to computer-enhanced learning. The most notable advantages of e-learning are flexibility, convenience and the ability to work at our own pace. Other advantages include the ability to communicate with fellow classmates from around the country. However, a common problem of usual e-learning approach is the lack of face-to-face interaction with a teacher. The most obvious solution to overcome this problem is to include a real-time multipoint audio-video channel to provide synchronous communications between teacher and students. This paper starts from an implementation experience of an e-learning platform for synchronous interaction between the actors, keeping "low-cost" as the main requirement of the deployment, to give the possibility to any user to participate at home using a domestic Internet access and a base-level personal computer. This paper proposes a software-based platform to provide multipoint communications in a heterogeneous unicast/multicast environment. It is constituted by a number of S-MCUs (software MCUs). Each S-MCU runs on a low-cost computer and is implemented like a software router by using click. Design, implementation and performance issues are discussed by a comparative numerical analysis.

Policy-based dynamic resource management may involve interaction between independent decision-making components which can lead to conflicts. For example, conflicts can occur between the policies for allocating resources and those setting quotas for users or classes of service. These policy conflicts cannot be detected by static analysis of the policies at specification-time as the conflicts arise from the current state of the resources within the system and so can only be detected at run-time. In this paper we use policies related to quality of service (QoS) provisioning for configuring differentiated services (DiffServ) networks to illustrate techniques for the dynamic detection and resolution of conflicts. Configuration includes implementing network provisioning decisions, performing admission control, and adapting bandwidth allocation dynamically according to emerging traffic demands. We identify possible conflicts between policies that manage the allocation of resources, and we also investigate conflicts that may arise between these policies and higher-level directives refined at the dynamic resource management level, acting as constraints. The paper shows how event calculus can be used to detect conflicts, focusing on the ones that emerge at run-time, and provides an approach for specifying policies to automate conflict resolution. The latter is demonstrated through our initial implementation of a dynamic conflict analysis tool

Policy-based management provides the ability to dynamically re-configure DiffServ networks such that desired Quality of Service (QoS) goals are achieved. This includes network provisioning decisions, performing admission control, and adapting bandwidth allocation dynamically. QoS management aims to satisfy the Service Level Agreements (SLAs) contracted by the provider and therefore QoS policies are derived from SLA specifications and the provider's business goals. This policy refinement is usually performed manually with no means of verifying that the policies written are supported by the network devices and actually achieve the desired QoS goals. Tool support is lacking and policy refinement has rarely been addressed in the literature. This paper extends our previous approach to policy refinement and shows how to apply it to the domain of DiffServ QoS management. We make use of goal elaboration and abductive reasoning to derive strategies that will achieve a given high-level goal. By combining these strategies with events and constraints, we show how policies can be refined, and what tool support can be provided for the refinement process using examples from the QoS management domain. The approach presented here can be used in other application domains such as storage area networks or security management.

Policy-based management provides the ability to dynamically re-configure DiffServ networks such that desired quality of service (QoS) goals are achieved. This includes network provisioning decisions, performing admission control, and adapting bandwidth allocation dynamically. QoS management aims to satisfy the service level agreements (SLAs) contracted by the provider and therefore QoS policies are derived from SLA specifications and the provider's business goals. This policy refinement is usually performed manually with no means of verifying that the policies written are supported by the network devices and actually achieve the desired QoS goals. Tool support is lacking and policy refinement has rarely been addressed in the literature. This paper extends our previous approach to policy refinement and shows how to apply it to the domain of DiffServ QoS management. We make use of goal elaboration and abductive reasoning to derive strategies that achieves a given high-level goal. By combining these strategies with events and constraints, we show how policies can be refined, and what tool support can be provided for the refinement process using examples from the QoS management domain. However, the approach presented here can be used in other application domains such as storage area networks or security management.

We argue that the evolution of requirements specifications can be supported by a cycle composed of two phases: analysis and revision. We investigate an instance of such a cycle, which combines two techniques of logical abduction and inductive learning to analyze and revise specifications respectively.