Author name cluster

Nathaniel D. Bastian

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers

1 author row

AAAI Conference 2026 Conference Paper

Consistency-based Abductive Reasoning over Perceptual Errors of Multiple Pre-trained Models in Novel Environments

Mario Leiva
Noel Ngu
Joshua Shay Kricheli
Aditya Taparia
Ransalu Senanayake
Paulo Shakarian
Nathaniel D. Bastian
John Corcoran

The deployment of pre-trained perception models in novel environments often leads to performance degradation due to distributional shifts. Although recent artificial intelligence approaches for metacognition use logical rules to characterize and filter model errors, improving precision often comes at the cost of reduced recall. This paper addresses the hypothesis that leveraging multiple pre-trained models can mitigate this recall reduction. We formulate the challenge of identifying and managing conflicting predictions from various models as a consistency-based abduction problem, building on the idea of abductive learning (ABL) but applying it to test-time instead of training. The input predictions and the learned error detection rules derived from each model are encoded in a logic program. We then seek an abductive explanation—a subset of model predictions—that maximizes prediction coverage while ensuring the rate of logical inconsistencies (derived from domain constraints) remains below a specified threshold. We propose two algorithms for this knowledge representation task: an exact method based on Integer Programming (IP) and an efficient Heuristic Search (HS). Through extensive experiments on a simulated aerial imagery dataset featuring controlled, complex distributional shifts, we demonstrate that our abduction-based framework outperforms individual models and standard ensemble baselines, achieving, for instance, average relative improvements of approximately 13.6% in F1-score and 16.6% in accuracy across 15 diverse test datasets when compared to the best individual model. Our results validate the use of consistency-based abduction as an effective mechanism to robustly integrate knowledge from multiple imperfect models in challenging, novel scenarios.

PDF Details DOI

IS Journal 2026 Journal Article

Metacognitive Artificial Intelligence in Vision Foundation Models: Research Challenges

Shahriar Rifat
A. Q. M. Sazzad Sayyed
Nathaniel D. Bastian
Francesco Restuccia

The adoption of Vision Foundation Models (VFMs) in high stakes scenarios has spurred the demand for task specific, high performance models. However, the lack of explainability of VFMs makes it challenging to ensure safety, reliability, and resilience across tasks when facing data distributions different from those seen during training. Recently, approaches based on metacognition, the human ability to regulate cognitive processes, have emerged as a promising way to understand these large models. This paper surveys the interdisciplinary connection between metacognition and state of the art VFMs, and further examines its relationship with knowledge distillation (KD), a widely used technique in VFMs. The paper concludes by outlining possible avenues for future research on the topic.

Details DOI

IS Journal 2026 Journal Article

Relevance Structures are a Necessary Prerequisite for Artificial Metacognition

Alexander M. Berenbeim
Ramneet Kaur
Adam Cobb
Brian Matejek
Anirban Roy
Susmit Jha
Nathaniel D. Bastian

AI-enabled systems deployed in dynamic, open-world, safety- and mission-critical environments often lack mechanisms for reliable self-monitoring and self-regulation under shift and perturbation. Artificial Metacognition (AM) addresses this by enabling systems to assess and adapt their own inference and learning processes. We show that relevance structures, which are order-theoretic substrates for comparing generalization and specialization and for quantifying under-coverage versus overreach, are structurally unavoidable for computable self-regulation. In turn, relevance structures provide an operational control theory for metacognitive tasks. We introduce exploratory, parsimonious, and balanced relevance scores and embed them into version-space algebra to regulate hypothesis search and inductive strategy selection. The resulting framework yields competence-aware learning systems that can both correct and explain their behavior by explicitly balancing exploration and parsimony. We illustrate relevance structures and their application with a medical visual question answering study on the SLAKE dataset.

Details DOI

AAAI Conference 2026 Conference Paper

SDE-HARL: Scalable Distributed Policy Execution for Heterogeneous-Agent Reinforcement Learning

Toan D. Gian
Mohammad Abdi
Nathaniel D. Bastian
Francesco Restuccia

HARL enables agents to execute cooperative tasks by adopting agent-specific policies. Most of existing HARL methods use individual policy neural networks to ensure monotonic improvement, which leads to substantial computational overhead. The proposed SDE-HARL overcomes this limitation by decomposing each agent's policy neural network into a lightweight local neural network and a global neural network executed at an edge server. Each local neural network generates and sends a compressed latent representation to the edge server, which aggregates the representations and produces agent-specific inferences. As such, SDE-HARL allows to significantly save computing and networking resources while preserving agent-specific behavior. A key feature of SDE-HARL is grouping agents with similar roles via a role-aware mechanism and share partial parameters in their global networks, while an identity-aware mechanism is introduced to promote behavioral diversity among agents within the same group. We prototyped SDE-HARL on an experimental testbed composed of a Jetson Nano and Raspberry PI to measure latency and network resource consumption. We evaluated SDE-HARL's performance on several benchmark datasets, including Google Research Football and StarCraft II. Experimental results show that SDE-HARL reaches up to 90% win rate while reducing latency, energy consumption, and networking overhead respectively by 2x, 2.5x, and 5x compared to existing work.

PDF Details DOI

EAAI Journal 2025 Journal Article

Data-efficient Federated Learning for Edge Network Intrusion Detection

David A. Bierbrauer
Sean M. Coffey
Mikal R. Willeke
John D. Beggs
Nathaniel D. Bastian

The use of deep learning enabled Network Intrusion Detection Systems (NIDS) has enabled greater efficiency in classifying and responding to cyber threats. As networks continue to grow and increasingly become interconnected, novel NIDS platforms need to be engineered to enhance data-efficiency and data privacy so that NIDS may share what they learned with each other. To accomplish this, we developed a Federated Learning (FL) NIDS platform where clients learn from raw network traffic payloads and share their model weights with a central server. This server aggregates the received weights and redistributes to each client. Our platform demonstrates the feasibility of this approach using both realistic edge hardware and simulation environments. Here, we examine an environment where each network has heterogeneous data distributions but must keep the training data locally. The proposed neural network architectures must also be relatively simple to enable learning at the edge on low-compute devices. Through experimentation, our federated NIDS platform powered by a simple convolutional neural network utilizing the FedProx FL aggregation algorithm achieved over 82% accuracy in a multi-class setting with a Cohen’s Kappa statistic of 0. 6853 after 60 rounds of FL.

Details DOI

TMLR Journal 2025 Journal Article

SAFE-NID: Self-Attention with Normalizing-Flow Encodings for Network Intrusion Detection

Brian Matejek
Ashish Gehani
Nathaniel D. Bastian
Daniel J Clouse
Bradford J Kline
Susmit Jha

Machine learning models are increasingly adopted to monitor network traffic and detect intrusions. In this work, we introduce SAFE-NID, a novel machine learning approach for real-time packet-level traffic monitoring and intrusion detection that includes a safeguard to detect zero day attacks as out-of-distribution inputs. Unlike traditional models, which falter against zero-day attacks and concept drift, SAFE-NID leverages a lightweight encoder-only transformer architecture combined with a novel normalizing flows-based safeguard. This safeguard not only quantifies uncertainty but also identifies out-of-distribution (OOD) inputs, enabling robust performance in dynamic threat landscapes. Our generative model learns class-conditional representations of the internal features of the deep neural network. We demonstrate the effectiveness of our approach by converting publicly available network flow-level intrusion datasets into packet-level ones. We release the labeled packet-level versions of these datasets with over 50 million packets each and describe the challenges in creating these datasets. We withhold from the training data certain attack categories to simulate zero-day attacks. Existing deep learning models, which achieve an accuracy of over 99% when detecting known attacks, only correctly classify 1% of the novel attacks. Our proposed transformer architecture with normalizing flows model safeguard achieves an area under the receiver operating characteristic curve of over 0.97 in detecting these novel inputs, outperforming existing combinations of neural architectures and model safeguards. The additional latency in processing each packet by the safeguard is a small fraction of the overall inference task. This dramatic improvement in detecting zero-day attacks and distribution shifts emphasizes SAFE-NID’s novelty and utility as a reliable and efficient safety monitoring tool for real-world network intrusion detection.

PDF Details

NeurIPS Conference 2024 Conference Paper

RGMDT: Return-Gap-Minimizing Decision Tree Extraction in Non-Euclidean Metric Space

Jingdi Chen
Hanhan Zhou
Yongsheng Mei
Carlee Joe-Wong
Gina Adam
Nathaniel D. Bastian
Tian Lan

Deep Reinforcement Learning (DRL) algorithms have achieved great success in solving many challenging tasks while their black-box nature hinders interpretability and real-world applicability, making it difficult for human experts to interpret and understand DRL policies. Existing works on interpretable reinforcement learning have shown promise in extracting decision tree (DT) based policies from DRL policies with most focus on the single-agent settings while prior attempts to introduce DT policies in multi-agent scenarios mainly focus on heuristic designs which do not provide any quantitative guarantees on the expected return. In this paper, we establish an upper bound on the return gap between the oracle expert policy and an optimal decision tree policy. This enables us to recast the DT extraction problem into a novel non-euclidean clustering problem over the local observation and action values space of each agent, with action values as cluster labels and the upper bound on the return gap as clustering loss. Both the algorithm and the upper bound are extended to multi-agent decentralized DT extractions by an iteratively-grow-DT procedure guided by an action-value function conditioned on the current DTs of other agents. Further, we propose the Return-Gap-Minimization Decision Tree (RGMDT) algorithm, which is a surprisingly simple design and is integrated with reinforcement learning through the utilization of a novel Regularized Information Maximization loss. Evaluations on tasks like D4RL show that RGMDT significantly outperforms heuristic DT-based baselines and can achieve nearly optimal returns under given DT complexity constraints (e. g. , maximum number of DT nodes).

PDF Details DOI

EAAI Journal 2023 Journal Article

SeNet-I: An approach for detecting network intrusions through serialized network traffic images

Yasir Ali Farrukh
Syed Wali
Irfan Khan
Nathaniel D. Bastian

The exponential growth of the internet and inter-connectivity has resulted in an extensive increase in network size and the corresponding data, which has led to numerous novel attacks that pose significant challenges to network security. However, conventional network security approaches predominantly rely on the metadata of network traffic, utilized in numeric form, which is becoming ineffective against new attacks that hide within the content of the traffic. Therefore, it raises the need for security systems to adapt to the changing dynamics of network attacks. To address this issue, we propose a new approach called SeNet-I that leverages computer vision capabilities to combine low-level features and develop a more abstract and high-level representation of network traffic without requiring feature engineering. The proposed approach utilizes the raw network traffic information and transforms it into serialized three-channel images, which are employed as input to a proposed deep concatenated convolutional neural network model. Additionally, SeNet-I can easily incorporate packet level information, which is often challenging for conventional approaches due to its high dimensionality. To demonstrate the effectiveness of the proposed approach, we tested SeNet-I on both packet-based and flow-based network traffic, comparing it with current state-of-the-art methods and different image-based approaches. With F1 scores of 96% and 83% achieved in the multi-class classification of flow-based and packet-based network intrusion detection, our proposed approach outperformed other existing methods in the literature. Lastly, we discussed the advantages and limitations of the proposed method.

Details DOI