arXiv subjects

Marta Kwiatkowska

Publications and source records attributed to Marta Kwiatkowska.

At least 55 records · Page 3Linked to original sources

arXiv2023

Certifiers Make Neural Networks Vulnerable to Availability Attacks

To achieve reliable, robust, and safe AI systems, it is vital to implement fallback strategies when AI predictions cannot be trusted. Certifiers for neural networks are a reliable way to check the robustness of these predictions. They guarantee for some predictions that a certain class of manipulations or attacks could not have changed the outcome. For the remaining predictions without guarantees, the method abstains from making a prediction, and a fallback strategy needs to be invoked, which typically incurs additional costs, can require a human operator, or even fail to provide any prediction. While this is a key concept towards safe and secure AI, we show for the first time that this approach comes with its own security risks, as such fallback strategies can be deliberately triggered by an adversary. In addition to naturally occurring abstains for some inputs and perturbations, the adversary can use training-time attacks to deliberately trigger the fallback with high probability. This transfers the main system load onto the fallback, reducing the overall system's integrity and/or availability. We design two novel availability attacks, which show the practical relevance of these threats. For example, adding 1% poisoned data during training is sufficient to trigger the fallback and hence make the model unavailable for up to 100% of all inputs by inserting the trigger. Our extensive experiments across multiple datasets, model architectures, and certifiers demonstrate the broad applicability of these attacks. An initial investigation into potential defenses shows that current approaches are insufficient to mitigate the issue, highlighting the need for new, specific solutions.

cs.LG↗

arXiv2023

Trust-Aware Motion Planning for Human-Robot Collaboration under Distribution Temporal Logic Specifications

Recent work has considered trust-aware decision making for human-robot collaboration (HRC) with a focus on model learning. In this paper, we are interested in enabling the HRC system to complete complex tasks specified using temporal logic that involve human trust. Since human trust in robots is not observable, we adopt the widely used partially observable Markov decision process (POMDP) framework for modelling the interactions between humans and robots. To specify the desired behaviour, we propose to use syntactically co-safe linear distribution temporal logic (scLDTL), a logic that is defined over predicates of states as well as belief states of partially observable systems. The incorporation of belief predicates in scLDTL enhances its expressiveness while simultaneously introducing added complexity. This also presents a new challenge as the belief predicates must be evaluated over the continuous (infinite) belief space. To address this challenge, we present an algorithm for solving the optimal policy synthesis problem. First, we enhance the belief MDP (derived by reformulating the POMDP) with a probabilistic labelling function. Then a product belief MDP is constructed between the probabilistically labelled belief MDP and the automaton translation of the scLDTL formula. Finally, we show that the optimal policy can be obtained by leveraging existing point-based value iteration algorithms with essential modifications. Human subject experiments with 21 participants on a driving simulator demonstrate the effectiveness of the proposed approach.

cs.RO↗

arXiv2023

When to Trust AI: Advances and Challenges for Certification of Neural Networks

Artificial intelligence (AI) has been advancing at a fast pace and it is now poised for deployment in a wide range of applications, such as autonomous systems, medical diagnosis and natural language processing. Early adoption of AI technology for real-world applications has not been without problems, particularly for neural networks, which may be unstable and susceptible to adversarial examples. In the longer term, appropriate safety assurance techniques need to be developed to reduce potential harm due to avoidable system failures and ensure trustworthiness. Focusing on certification and explainability, this paper provides an overview of techniques that have been developed to ensure safety of AI decisions and discusses future challenges.

cs.LG↗

arXiv2023

When are Local Queries Useful for Robust Learning?

Distributional assumptions have been shown to be necessary for the robust learnability of concept classes when considering the exact-in-the-ball robust risk and access to random examples by Gourdeau et al. (2019). In this paper, we study learning models where the learner is given more power through the use of local queries, and give the first distribution-free algorithms that perform robust empirical risk minimization (ERM) for this notion of robustness. The first learning model we consider uses local membership queries (LMQ), where the learner can query the label of points near the training sample. We show that, under the uniform distribution, LMQs do not increase the robustness threshold of conjunctions and any superclass, e.g., decision lists and halfspaces. Faced with this negative result, we introduce the local equivalence query ($\mathsf{LEQ}$) oracle, which returns whether the hypothesis and target concept agree in the perturbation region around a point in the training sample, as well as a counterexample if it exists. We show a separation result: on the one hand, if the query radius $λ$ is strictly smaller than the adversary's perturbation budget $ρ$, then distribution-free robust learning is impossible for a wide variety of concept classes; on the other hand, the setting $λ=ρ$ allows us to develop robust ERM algorithms. We then bound the query complexity of these algorithms based on online learning guarantees and further improve these bounds for the special case of conjunctions. We finish by giving robust learning algorithms for halfspaces on $\{0,1\}^n$ and then obtaining robustness guarantees for halfspaces in $\mathbb{R}^n$ against precision-bounded adversaries.

cs.LG↗

arXiv2023

Adversarial Robustness Certification for Bayesian Neural Networks

We study the problem of certifying the robustness of Bayesian neural networks (BNNs) to adversarial input perturbations. Given a compact set of input points $T \subseteq \mathbb{R}^m$ and a set of output points $S \subseteq \mathbb{R}^n$, we define two notions of robustness for BNNs in an adversarial setting: probabilistic robustness and decision robustness. Probabilistic robustness is the probability that for all points in $T$ the output of a BNN sampled from the posterior is in $S$. On the other hand, decision robustness considers the optimal decision of a BNN and checks if for all points in $T$ the optimal decision of the BNN for a given loss function lies within the output set $S$. Although exact computation of these robustness properties is challenging due to the probabilistic and non-convex nature of BNNs, we present a unified computational framework for efficiently and formally bounding them. Our approach is based on weight interval sampling, integration, and bound propagation techniques, and can be applied to BNNs with a large number of parameters, and independently of the (approximate) inference method employed to train the BNN. We evaluate the effectiveness of our methods on various regression and classification tasks, including an industrial regression benchmark, MNIST, traffic sign recognition, and airborne collision avoidance, and demonstrate that our approach enables certification of robustness and uncertainty of BNN predictions.

cs.LG↗

arXiv2023

Emergent Linguistic Structures in Neural Networks are Fragile

Large Language Models (LLMs) have been reported to have strong performance on natural language processing tasks. However, performance metrics such as accuracy do not measure the quality of the model in terms of its ability to robustly represent complex linguistic structures. In this paper, focusing on the ability of language models to represent syntax, we propose a framework to assess the consistency and robustness of linguistic representations. To this end, we introduce measures of robustness of neural network models that leverage recent advances in extracting linguistic constructs from LLMs via probing tasks, i.e., simple tasks used to extract meaningful information about a single facet of a language model, such as syntax reconstruction and root identification. Empirically, we study the performance of four LLMs across six different corpora on the proposed robustness measures by analysing their performance and robustness with respect to syntax-preserving perturbations. We provide evidence that context-free representation (e.g., GloVe) are in some cases competitive with context-dependent representations from modern LLMs (e.g., BERT), yet equally brittle to syntax-preserving perturbations. Our key observation is that emergent syntactic representations in neural networks are brittle. We make the code, trained models and logs available to the community as a contribution to the debate about the capabilities of LLMs.

cs.LG↗

arXiv2023

Sample Efficient Model-free Reinforcement Learning from LTL Specifications with Optimality Guarantees

Linear Temporal Logic (LTL) is widely used to specify high-level objectives for system policies, and it is highly desirable for autonomous systems to learn the optimal policy with respect to such specifications. However, learning the optimal policy from LTL specifications is not trivial. We present a model-free Reinforcement Learning (RL) approach that efficiently learns an optimal policy for an unknown stochastic system, modelled using Markov Decision Processes (MDPs). We propose a novel and more general product MDP, reward structure and discounting mechanism that, when applied in conjunction with off-the-shelf model-free RL algorithms, efficiently learn the optimal policy that maximizes the probability of satisfying a given LTL specification with optimality guarantees. We also provide improved theoretical results on choosing the key parameters in RL to ensure optimality. To directly evaluate the learned policy, we adopt probabilistic model checker PRISM to compute the probability of the policy satisfying such specifications. Several experiments on various tabular MDP environments across different LTL tasks demonstrate the improved sample efficiency and optimal policy convergence.

cs.LG↗

arXiv2023

Compositional Probabilistic and Causal Inference using Tractable Circuit Models

Probabilistic circuits (PCs) are a class of tractable probabilistic models, which admit efficient inference routines depending on their structural properties. In this paper, we introduce md-vtrees, a novel structural formulation of (marginal) determinism in structured decomposable PCs, which generalizes previously proposed classes such as probabilistic sentential decision diagrams. Crucially, we show how mdvtrees can be used to derive tractability conditions and efficient algorithms for advanced inference queries expressed as arbitrary compositions of basic probabilistic operations, such as marginalization, multiplication and reciprocals, in a sound and generalizable manner. In particular, we derive the first polytime algorithms for causal inference queries such as backdoor adjustment on PCs. As a practical instantiation of the framework, we propose MDNets, a novel PC architecture using md-vtrees, and empirically demonstrate their application to causal inference.

cs.AI↗

arXiv2022

Bayesian Network Models of Causal Interventions in Healthcare Decision Making: Literature Review and Software Evaluation

This report summarises the outcomes of a systematic literature search to identify Bayesian network models used to support decision making in healthcare. After describing the search methodology, the selected research papers are briefly reviewed, with the view to identify publicly available models and datasets that are well suited to analysis using the causal interventional analysis software tool developed in Wang B, Lyle C, Kwiatkowska M (2021). Finally, an experimental evaluation of applying the software on a selection of models is carried out and preliminary results are reported.

cs.AI↗

arXiv2022

Symbolic Verification and Strategy Synthesis for Turn-based Stochastic Games

Stochastic games are a convenient formalism for modelling systems that comprise rational agents competing or collaborating within uncertain environments. Probabilistic model checking techniques for this class of models allow us to formally specify quantitative specifications of either collective or individual behaviour and then automatically synthesise strategies for the agents under which these specifications are guaranteed to be satisfied. Although good progress has been made on algorithms and tool support, efficiency and scalability remain a challenge. In this paper, we investigate a symbolic implementation based on multi-terminal binary decision diagrams. We describe how to build and verify turn-based stochastic games against either zero-sum or Nash equilibrium based temporal logic specifications. We collate a set of benchmarks for this class of games, and evaluate the performance of our approach, showing that it is superior in a number of cases and that strategies synthesised in a symbolic fashion can be considerably more compact.

cs.LO↗

arXiv2022

Robustness of Unsupervised Representation Learning without Labels

Unsupervised representation learning leverages large unlabeled datasets and is competitive with supervised learning. But non-robust encoders may affect downstream task robustness. Recently, robust representation encoders have become of interest. Still, all prior work evaluates robustness using a downstream classification task. Instead, we propose a family of unsupervised robustness measures, which are model- and task-agnostic and label-free. We benchmark state-of-the-art representation encoders and show that none dominates the rest. We offer unsupervised extensions to the FGSM and PGD attacks. When used in adversarial training, they improve most unsupervised robustness measures, including certified robustness. We validate our results against a linear probe and show that, for MOCOv2, adversarial training results in 3 times higher certified accuracy, a 2-fold decrease in impersonation attack success rate and considerable improvements in certified robustness.

cs.LG↗

arXiv2022

Tractable Uncertainty for Structure Learning

Bayesian structure learning allows one to capture uncertainty over the causal directed acyclic graph (DAG) responsible for generating given data. In this work, we present Tractable Uncertainty for STructure learning (TRUST), a framework for approximate posterior inference that relies on probabilistic circuits as the representation of our posterior belief. In contrast to sample-based posterior approximations, our representation can capture a much richer space of DAGs, while also being able to tractably reason about the uncertainty through a range of useful inference queries. We empirically show how probabilistic circuits can be used as an augmented representation for structure learning methods, leading to improvement in both the quality of inferred structures and posterior uncertainty. Experimental results on conditional query answering further demonstrate the practical utility of the representational capacity of TRUST.

cs.LG↗

arXiv2022

Probabilistic Model Checking for Strategic Equilibria-based Decision Making: Advances and Challenges

Game-theoretic concepts have been extensively studied in economics to provide insight into competitive behaviour and strategic decision making. As computing systems increasingly involve concurrently acting autonomous agents, game-theoretic approaches are becoming widespread in computer science as a faithful modelling abstraction. These techniques can be used to reason about the competitive or collaborative behaviour of multiple rational agents with distinct goals or objectives. This paper provides an overview of recent advances in developing a modelling, verification and strategy synthesis framework for concurrent stochastic games implemented in the probabilistic model checker PRISM-games. This is based on a temporal logic that supports finite- and infinite-horizon temporal properties in both a zero-sum and nonzero-sum setting, the latter using Nash and correlated equilibria with respect to two optimality criteria, social welfare and social fairness. We summarise the key concepts, logics and algorithms and the currently available tool support. Future challenges and recent progress in adapting the framework and algorithmic solutions to continuous environments and neural networks are also outlined.

cs.FL↗

arXiv2022

Finite-horizon Equilibria for Neuro-symbolic Concurrent Stochastic Games

We present novel techniques for neuro-symbolic concurrent stochastic games, a recently proposed modelling formalism to represent a set of probabilistic agents operating in a continuous-space environment using a combination of neural network based perception mechanisms and traditional symbolic methods. To date, only zero-sum variants of the model were studied, which is too restrictive when agents have distinct objectives. We formalise notions of equilibria for these models and present algorithms to synthesise them. Focusing on the finite-horizon setting, and (global) social welfare subgame-perfect optimality, we consider two distinct types: Nash equilibria and correlated equilibria. We first show that an exact solution based on backward induction may yield arbitrarily bad equilibria. We then propose an approximation algorithm called frozen subgame improvement, which proceeds through iterative solution of nonlinear programs. We develop a prototype implementation and demonstrate the benefits of our approach on two case studies: an automated car-parking system and an aircraft collision avoidance system.

cs.GT↗

arXiv2022

Learning Dynamics and Generalization in Reinforcement Learning

Solving a reinforcement learning (RL) problem poses two competing challenges: fitting a potentially discontinuous value function, and generalizing well to new observations. In this paper, we analyze the learning dynamics of temporal difference algorithms to gain novel insight into the tension between these two objectives. We show theoretically that temporal difference learning encourages agents to fit non-smooth components of the value function early in training, and at the same time induces the second-order effect of discouraging generalization. We corroborate these findings in deep RL agents trained on a range of environments, finding that neural networks trained using temporal difference algorithms on dense reward tasks exhibit weaker generalization between states than randomly initialized networks and networks trained with policy gradient methods. Finally, we investigate how post-training policy distillation may avoid this pitfall, and show that this approach improves generalization to novel environments in the ProcGen suite and improves robustness to input perturbations.

cs.LG↗

arXiv2022

Sample Complexity Bounds for Robustly Learning Decision Lists against Evasion Attacks

A fundamental problem in adversarial machine learning is to quantify how much training data is needed in the presence of evasion attacks. In this paper we address this issue within the framework of PAC learning, focusing on the class of decision lists. Given that distributional assumptions are essential in the adversarial setting, we work with probability distributions on the input data that satisfy a Lipschitz condition: nearby points have similar probability. Our key results illustrate that the adversary's budget (that is, the number of bits it can perturb on each input) is a fundamental quantity in determining the sample complexity of robust learning. Our first main result is a sample-complexity lower bound: the class of monotone conjunctions (essentially the simplest non-trivial hypothesis class on the Boolean hypercube) and any superclass has sample complexity at least exponential in the adversary's budget. Our second main result is a corresponding upper bound: for every fixed $k$ the class of $k$-decision lists has polynomial sample complexity against a $\log(n)$-bounded adversary. This sheds further light on the question of whether an efficient PAC learning algorithm can always be used as an efficient $\log(n)$-robust learning algorithm under the uniform distribution.

cs.LG↗

arXiv2022

Individual Fairness Guarantees for Neural Networks

We consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). In particular, we work with the $ε$-$δ$-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of $ε$-similar individuals is bounded by a maximum decision tolerance $δ\geq 0$. Working with a range of metrics, including the Mahalanobis distance, we propose a method to overapproximate the resulting optimisation problem using piecewise-linear functions to lower and upper bound the NN's non-linearities globally over the input space. We encode this computation as the solution of a Mixed-Integer Linear Programming problem and demonstrate that it can be used to compute IF guarantees on four datasets widely used for fairness benchmarking. We show how this formulation can be used to encourage models' fairness at training time by modifying the NN loss, and empirically confirm our approach yields NNs that are orders of magnitude fairer than state-of-the-art methods.

cs.LG↗

arXiv2022

Robustness Guarantees for Credal Bayesian Networks via Constraint Relaxation over Probabilistic Circuits

In many domains, worst-case guarantees on the performance (e.g., prediction accuracy) of a decision function subject to distributional shifts and uncertainty about the environment are crucial. In this work we develop a method to quantify the robustness of decision functions with respect to credal Bayesian networks, formal parametric models of the environment where uncertainty is expressed through credal sets on the parameters. In particular, we address the maximum marginal probability (MARmax) problem, that is, determining the greatest probability of an event (such as misclassification) obtainable for parameters in the credal set. We develop a method to faithfully transfer the problem into a constrained optimization problem on a probabilistic circuit. By performing a simple constraint relaxation, we show how to obtain a guaranteed upper bound on MARmax in linear time in the size of the circuit. We further theoretically characterize this constraint relaxation in terms of the original Bayesian network structure, which yields insight into the tightness of the bound. We implement the method and provide experimental evidence that the upper bound is often near tight and demonstrates improved scalability compared to other methods.

cs.AI↗