Dawood Wasif

Federated learning enables on-device training without centralizing data, yet existing systems still struggle to provide explanations that are both locally faithful and globally consistent under strict privacy and bandwidth constraints. Prior approaches either keep explanations siloed across clients, transmit heavy or sensitive artifacts, or replace expressive task models with interpretable surrogates that sacrifice accuracy. We propose xFedAlign, a model-agnostic framework that decouples task optimization in parameter space from explanation coordination in a compact group space. Each client distills a lightweight surrogate to produce private, per-class top-k attribution artifacts, which are robustly aggregated by the server into a Global Explanation Prior that softly aligns client explanations without constraining task learning. Across image, text, and tabular benchmarks with IID and non-IID partitions, xFedAlign matches FedAvg accuracy while consistently reducing explanation drift and improving deletion and insertion AUC relative to Local-XAI, FedAttr-Agg, and Fed-XAI, with only a few kilobytes of additional communication per round. Privacy and robustness evaluations further demonstrate reduced membership inference advantage and increased resistance to attribution poisoning, enabling consistent and trustworthy explanations in federated learning.

@inproceedings{wasif2026xfedalign,
  title={Explainable Federated Learning via Global--Local Attribution Alignment},
  author={Wasif, Dawood and Moore, Terrence J. and Lu, Chang-Tien and Cho, Jin-Hee},
  booktitle={Proceedings of the International Conference on Machine Learning},
  year={2026},
  url={https://openreview.net/forum?id=gicCGXeG2P}
}

Federated Learning (FL) has gained prominence in machine learning applications across critical domains, offering collaborative model training without centralized data aggregation. However, FL frameworks that protect privacy often sacrifice fairness and reliability; differential privacy reduces data leakage but hides sensitive attributes needed for bias correction, worsening performance gaps across demographic groups. This work explores the trade-off between privacy and fairness in FL-based object detection and introduces RESFL, an integrated solution optimizing both. RESFL incorporates adversarial privacy disentanglement and uncertainty-guided fairness-aware aggregation. The adversarial component uses a gradient reversal layer to remove sensitive attributes, reducing privacy risks while maintaining fairness. The uncertainty-aware aggregation employs an evidential neural network to weight client updates adaptively, prioritizing contributions with lower fairness disparities and higher confidence. This ensures robust and equitable FL model updates. We demonstrate the effectiveness of RESFL in high-stakes autonomous vehicle scenarios, where it achieves high mAP on FACET and CARLA, reduces membership-inference attack success by 37%, reduces equality-of-opportunity gap by 17% relative to the FedAvg baseline, and maintains superior adversarial robustness. However, RESFL is inherently domain-agnostic and thus applicable to a broad range of application domains beyond autonomous driving.

@inproceedings{wasif2026resfl,
  title={RESFL: An Uncertainty-Aware Framework for Responsible Federated Learning by Balancing Privacy, Fairness and Utility},
  author={Wasif, Dawood and Moore, Terrence J. and Cho, Jin-Hee},
  booktitle={International Conference on Learning Representations},
  year={2026},
  url={https://arxiv.org/abs/2503.16251}
}

When algorithmic decisions depend on data distributed across institutions, how can we ensure that an individual’s outcome does not change arbitrarily based on a protected attribute? We study this question in vertical federated learning (VFL), where features are split across parties, sensitive attributes may be private, and proxies for protected characteristics can be scattered across institutional boundaries under strict privacy constraints. Our focus is on individual-level counterfactual stability, i.e., per-instance prediction consistency under protected-attribute interventions as formalized in the causal fairness literature, rather than group parity guarantees such as demographic parity or equalized odds. We propose SCC-VFL, a server-centric framework for enforcing selective counterfactual consistency (SCC) at the individual level in VFL. SCC-VFL operationalizes a given policy specification by combining three components: (i) differentially private, graph-free discovery of feature roles into non-descendants, policy-permitted mediators, and impermissible proxies using only a formally private sketch of the sensitive attribute, with a formal per-release privacy that does not extend to the full training pipeline; (ii) masked counterfactual generation that edits only mediators while fixing non-descendants and suppressing proxy leakage; and (iii) server-side enforcement via an SCC consistency loss that penalizes impermissible prediction changes under protected-attribute interventions. Across three real-world datasets spanning credit, healthcare, and criminal justice, SCC-VFL maintains or improves predictive accuracy while sharply reducing decision flip rates by up to 98% relative to strong baselines. It also lowers attribute-inference attack success and improves robustness, demonstrating favorable utility-fairness-privacy trade-offs in realistic VFL deployments.

@inproceedings{wasif2026scc,
  title={Toward Individual Fairness Without Centralized Data: Selective Counterfactual Consistency for Vertical Federated Learning},
  author={Wasif, Dawood and Reddy, Chandan K. and Moore, Terrence J. and Cho, Jin-Hee},
  booktitle={Proceedings of the ACM Conference on Fairness, Accountability, and Transparency},
  year={2026},
  url={https://openreview.net/forum?id=OjD3qbcRwr}
}

Coordinated drone swarms offer transformative capabilities in defense, surveillance, and disaster response but pose core AI alignment challenges: ensuring safe, corrigible, and robust behavior under uncertainty and adversarial threat. Existing human-in-the-loop and shared-autonomy approaches often rely on fixed trust models or centralized overrides, lack principled mechanisms for uncertainty estimation, and fail to scale securely. We present SHIELD, a unified framework for secure, scalable human–swarm teaming that integrates: (i) evidential deep learning to quantify epistemic uncertainty via vacuity scores; (ii) a dynamic trust graph that filters compromised agents through behavior-based trust propagation; and (iii) a lightweight intrusion response system for real-time attack detection, quarantine, and recovery. Human oversight is triggered only when agents encounter unfamiliar inputs or consensus breaks down, minimizing cognitive load. In high-fidelity simulations, SHIELD reduces collision penalties by 10%, improves mission completion by 25%, and detects over 80% of GPS spoofing, jamming, and replay attacks while halving human interventions. Ablation studies confirm each module’s importance, and scalability experiments show graceful performance degradation from 8 to 32 agents and increasing obstacle density. Our findings demonstrate that SHIELD’s integration of uncertainty, trust, and embedded defense offers a principled and practical solution to real-world alignment in multi-agent autonomous systems.

@inproceedings{wasif2026shield,
  title={SHIELD: Secure Human-Machine Interaction with Evidential Learning and Dynamic Trust for Drone Swarm Control},
  author={Wasif, Dawood and Moore, Terrence J. and Yoon, Seunghyun and Lim, Hyuk and Kim, Dan and Nelson, Frederica F. and Cho, Jin-Hee},
  booktitle={Proceedings of the USENIX Workshop on Autonomous Systems Security},
  year={2026},
  url={https://arxiv.org/pdf/2605.07117v1}
}

Autonomous vehicles must remain safe and effective when encountering rare long-tailed scenarios or cyber–physical intrusions during driving. We propose a risk-aware human-in-the-loop framework, RAIL, that explicitly closes the detect–respond–learn loop within a unified runtime architecture. RAIL introduces an interpretable Intrusion Risk Index (IRI) that fuses heterogeneous runtime cues, including curvature actuation integrity, time-to-collision proximity, and observation-shift consistency, through a weighted Noisy-OR to enable probabilistic, source-decomposed risk quantification at control rate. When risk rises, a contextual bandit selects cue-specific shields and a graded authority blending mechanism interpolates between nominal and safeguarded actions, preserving efficiency under low risk while enabling proactive containment under threat. RAIL further integrates dual reward shaping and risk-prioritized replay so that takeovers and near misses steer learning. Across MetaDrive and CARLA, RAIL matches expert-level success while reducing safety violations and operator interventions relative to RL, safe RL, offline and imitation baselines. Under Controller Area Network (CAN) injection and LiDAR spoofing, RAIL reduces attack success rates by up to 48% and lowers disengagement under attack by more than 50%, demonstrating that explicit multi-source risk modeling yields substantial robustness gains in safety-critical autonomy.

@inproceedings{wasif2026rail,
  title={Risk-Aware Human-in-the-Loop Framework with Adaptive Intrusion Response for Autonomous Vehicles},
  author={Wasif, Dawood and Moore, Terrence J. and Yoon, Seunghyun and Lim, Hyuk and Kim, Dan and Nelson, Frederica F. and Cho, Jin-Hee},
  booktitle={Proceedings of the USENIX Workshop on Autonomous Systems Security},
  year={2026}
}

The increasing complexity of cloud-native infrastructure has made Infrastructure-as-Code (IaC) essential for reproducible and scalable deployments. While large language models (LLMs) have shown promise in generating IaC snippets from natural language prompts, their monolithic, single-pass generation approach often results in syntactic errors, policy violations, and unscalable designs. In this paper, we propose MACOG (Multi-Agent Code-Orchestrated Generation), a novel multi-agent LLM-based architecture for IaC generation that decomposes the task into modular subtasks handled by specialized agents: Architect, Provider Harmonizer, Engineer, Reviewer, Security Prover, Cost and Capacity Planner, DevOps, and Memory Curator. The agents interact via a shared-blackboard, finite-state orchestrator layer, and collectively produce Terraform configurations that are not only syntactically valid but also policy-compliant and semantically coherent. To ensure infrastructure correctness and governance, we incorporate Terraform Plan for execution validation and Open Policy Agent (OPA) for customizable policy enforcement. We evaluate MACOG using the IaC-Eval benchmark, where MACOG is the top enhancement across models, e.g., GPT-5 improves from 54.90 (RAG) to 74.02 and Gemini-2.5 Pro from 43.56 to 60.13, with concurrent gains on BLEU, CodeBERTScore, and an LLM-judge metric. Ablations show constrained decoding and deploy feedback are critical: removing them drops IaC-Eval to 64.89 and 56.93, respectively.

@inproceedings{wasif2026macog,
  title={Multi-Agent Code-Orchestrated Generation for Reliable Infrastructure-as-Code},
  author={Khan, Rana Nameer Hussain and Wasif, Dawood and Cho, Jin-Hee and Butt, Ali R.},
  booktitle={Proceedings of the Cloud Computing Conference},
  year={2026}
}