Publications

  1. David Fernandez, Pedram MohajerAnsari, Cigdem Kokenoz, Amir Salarpour, Bing Li, Mert D. Pese (2026)
    From MIRAGE to CLEAR: Component-Level Explainable Anomaly Reasoning for Autonomous Vehicle Perception Systems

    Uses language models to identify which vehicle perception components failed and explain the failures using traffic rules.

    IEEE/IFIP International Conference on Dependable Systems and Networks (DSN 2026)
    PDF Slides Code DOI Cite
  2. David Fernandez, Pedram MohajerAnsari, Amir Salarpour, Long Cheng, Abolfazl Razi, Mert D. Pesé (2026)
    Comparative Analysis of Patch Attack on VLM-Based Autonomous Driving Architectures

    Benchmarks vision-language driving systems against physical patch attacks, revealing how their security weaknesses differ.

    2026 IEEE Intelligent Vehicles Symposium (IV)
    PDF Slides Code DOI Cite
  3. Pedram MohajerAnsari, Amir Salarpour, David Fernandez, Mert D. Pesé (2026)
    Budget-Aware Adaptive Adversarial Patches for Black-Box Object Detection

    Optimizes adversarial patches under limited queries to expose object-detector vulnerabilities without access to model internals.

    2026 IEEE International Conference on Image Processing (ICIP) — Accepted
    PDF Slides (PPTX) DOI Cite
  4. Pedram MohajerAnsari, Amir Salarpour, David Fernandez, Cigdem Kokenoz, Bing Li, Mert D. Pesé (2025)
    Attention-Aware Temporal Adversarial Shadows on Traffic Sign Sequences

    Uses model attention to optimize realistic shadow attacks across video frames, exposing weaknesses in traffic sign recognition.

    IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2025), pp. 3600–3608
    PDF Code DOI Cite
  5. David Fernandez, Pedram MohajerAnsari, Cigdem Kokenoz, Amir Salarpour, Bing Li, Mert D. Pese (2025)
    WIP: From Detection to Explanation: Using LLMs for Adversarial Scenario Analysis in Vehicles

    Tests whether language models can distinguish perception errors from attacks and explain driving anomalies using traffic rules.

    3rd USENIX Symposium on Vehicle Security and Privacy (VehicleSec 2025)
    PDF Slides Code Cite
  6. Amir Salarpour, Pedram MohajerAnsari, David Fernandez, Mert D. Pese (2025)
    SASA: Sequence-Aware Shadow Attacks via Attention Alignment for Traffic Sign Recognition

    Creates realistic shadow attacks that fool traffic sign recognition across video frames without querying the target model.

    6th Workshop on Adversarial Machine Learning on Computer Vision: Safety of Vision-Language Agents (AdvML@CVPR)
    PDF Cite
  7. David Fernandez, Pedram MohajerAnsari, Amir Salarpour, Mert D. Pese (2026)
    Understanding Adversarial Transferability in Vision-Language Models for Autonomous Driving: A Cross-Architecture Analysis

    Measures how attacks transfer between vision-language driving systems, revealing security risks when the target model is unknown.

    SAE Technical Paper 2026-01-0170, WCX SAE World Congress Experience
    PDF Slides DOI Cite
  8. David Fernandez, Pedram MohajerAnsari, Amir Salarpour, Mert D. Pese (2025)
    Avoiding the Crash: A Vision-Language Model Evaluation of Critical Traffic Scenarios

    Evaluates vision-language models on real crash videos to measure whether their driving decisions could help prevent collisions.

    SAE Technical Paper 2025-01-8213, WCX SAE World Congress Experience
    Outstanding Student Paper
    PDF Slides DOI Cite
  9. Edward B. Duffy, David Fernandez, Alta de Waal, Mert D. Pesé (2025)
    Small Language Models on the Edge for Real-World Agentic Systems in Industry

    Combines small language models with agent workflows to run enterprise tasks locally, reducing latency and dependence on cloud services.

    Southern African Conference for Artificial Intelligence Research (SACAIR 2025)
    PDF Code Cite
  10. Philipp Bauerfeind, Amir Salarpour, David Fernandez, Pedram MohajerAnsari, Johannes Reschke, Mert D. Pesé (2025)
    David vs. Goliath: A Comparative Study of Different-Sized LLMs for Code Generation in the Domain of Automotive Scenario Generation

    Benchmarks language models for generating driving simulations, showing how retrieved examples improve smaller models and comparing quality across model sizes.

    arXiv preprint
    PDF DOI Cite