|
| |||||||||||||||
TPAD 2027 : Trustworthy Perception for Autonomous Driving: Adversarial Robustness, Multimodal Fusion, and Foundation Models | |||||||||||||||
| Link: https://tpad-2027.github.io/ | |||||||||||||||
| |||||||||||||||
Call For Papers | |||||||||||||||
|
TPAD 2027: Trustworthy Perception for Autonomous Driving: Adversarial Robustness, Multimodal Fusion, and Foundation Models
An Official Satellite Workshop of IEEE ICASSP 2027 (2027 IEEE International Conference on Acoustics, Speech, and Signal Processing) Date: May 17, 2027 Location: Toronto, Canada (Co-located with IEEE ICASSP 2027) ==================================================================== IMPORTANT DATES ==================================================================== - Paper Submission Deadline: December 2, 2026 (Extended) - Acceptance Notification: December 16, 2026 - Camera-Ready Paper Due: January 6, 2027 - Workshop Date: May 17, 2027 ==================================================================== OVERVIEW & SCOPE ==================================================================== Held as an official satellite workshop of IEEE ICASSP 2027, TPAD 2027 addresses trust in automated driving as a claim about failure bounds, not average accuracy. A trustworthy perception system keeps errors bounded, detectable at runtime, and traceable to evidence, so the motion planner can identify untrusted inputs. Cameras, LiDAR, and FMCW radar measure physical environments directly, allowing adversaries to manipulate system inputs via printed patches, light projections, spoofed radar chirps, or injected LiDAR returns. Perturbations propagate through calibration, data association, multi-modal fusion, and object tracking, while vision-language-action (VLA) models extend this attack surface across semantic boundaries. TPAD establishes an end-to-end signal processing and assurance perspective. To standardize evaluations, submissions must explicitly define four structural parameters: * Attacker Access: physical, digital, or semantic vector control. * Injection Point: the entry stage within the processing pipeline. * Stages Affected: downstream propagation across perception and tracking. * Measurement Point: sensor, perception, planner, or closed-loop trajectory level. ==================================================================== TOPICS OF INTEREST ==================================================================== - Physical-layer attacks: adversarial patches, optical interference, and light projection - FMCW and 4D imaging radar spoofing, jamming, and interference mitigation - LiDAR point-cloud injection, signal suppression, and relay attacks - Hardware vulnerabilities, clock desynchronization, and firmware-level manipulation - Robust multi-modal fusion across camera, LiDAR, and radar - Cross-modal vulnerability transfer and shared-representation attacks - Adversarial resilience of BEV, 3D occupancy, and vector-map representations - Sensor redundancy bounds and fail-operational degraded modes - Threat propagation across association, tracking, and trajectory prediction - Prompt injection, jailbreaking, and semantic attacks on VLA and world models - Physical-consistency verification, runtime monitoring, and spoof detection - Closed-loop safety-referred evaluation and hardware-in-the-loop threat testing ==================================================================== SUBMISSION & PUBLICATION ==================================================================== Submitted papers must follow the official IEEE ICASSP 2027 paper style, format, and length. Accepted papers will be published in the IEEE Xplore Digital Library as part of the ICASSP 2027 workshop proceedings and must be presented in person by an author. ==================================================================== ORGANIZING COMMITTEE ==================================================================== - Christos Korgialas (Aristotle University of Thessaloniki; NOEMAC PC, Greece) - Pai Chet Ng (Singapore Institute of Technology, Singapore) - Xiaoxiao Miao (Duke Kunshan University, China) - Arash Mohammadi (Concordia University, Canada) - Konstantinos N. Plataniotis - Scientific Advisor (University of Toronto, Canada) Contact: ckorgial@csd.auth.gr |
|