Home
The “2026 IEEE SPS Seasonal School on Secure and Trustworthy AI for Audio-Visual Media (SMART-AIM'26)” will take place on 22-24 September 2026 at Taxila, Pakistan, an ancient city just 30 Km from the Capital city Islamabad. Taxila is a city of the Gandharan civilization, sometimes known as one of its capitals, whose history can be traced from early microlithic communities at the Khanpur caves up to almost 1000 CE. Taxila was a hub of Buddhism, a centre of learning, an urban metropolis and a meeting point of various cultures, namely the Achaemenids, Greeks, Mauryans, Scythians, Parthians, Kushans, Huns and eventually the Muslims.
The school will be held at University of Engineering & Technology, Taxila in collaboration with Swarm Robotics Lab under National Centre of Robotics and Automation (NCRA).
This seasonal school aims to equip participants with cutting-edge knowledge and practical skills to address these challenges. The program will focus on secure AI-driven frameworks for detecting and mitigating audio-visual spoofing attacks, emphasizing both theoretical foundations and hands-on implementation. Participants will explore state-of-the-art approaches in deep-fake detection, including convolutional neural networks, vision transformers, and self-supervised learning models tailored for audio and video domains. Special attention will be given to federated learning strategies that enable privacy-preserving authentication, as well as lightweight architectures optimized for edge devices.
The school will feature expert-led technical talks, interactive workshop, and industry experience showcases, creating a platform for knowledge exchange between academia, industry, and regulatory bodies. Attendees will gain insights into dataset development for multilingual and low-resource languages, adversarial robustness, and ethical considerations in deploying anti-spoofing technologies. By fostering collaboration and innovation, this initiative seeks to build capacity in designing secure, trustworthy AI systems that can withstand evolving threats in the era of synthetic media.
We encourage participation of a spectrum of faculty members, researchers, including students and industry practitioners. Attendees will also benefit from networking opportunities in a friendly and stimulating environment. This event will also provide an opportunity to academia and industry to bridge the gap between academic research and industry ventures.
The school will focus on topics (but not limited too):
- Fundamentals of Deepfake Technology and Emerging Threats
- Generative AI Models: GANs, VAEs, and Diffusion Networks
- Audio Spoofing Detection: Spectral Features and ML/DL Approaches
- Self-Supervised Learning for Multilingual Anti-Spoofing
- Vision Transformers and Hybrid Architectures for Video Deepfake Detection
- Local–Global Feature Fusion for Robust Cross-Domain Detection
- Federated Learning for Privacy-Preserving Authentication
- Lightweight Anti-Spoofing Models for Edge and Mobile Devices
- Dataset Development for Low-Resource and Code-Switching Scenarios
- Adversarial Attacks and Robustness in PAD Systems
- Ethical, Legal, and Regulatory Frameworks for Secure AI
- Industry Applications: Banking, FinTech, and Digital Identity Solutions
- Practical use cases and integration of secure AI in real-world systems.
