Research Milestone: 3 Papers Accepted to IEEE ICASSP 2026 & Meet Us in Barcelona

Febuary 04, 2026

We’re excited to share that our research group has had three papers accepted to ICASSP 2026 (Barcelona, Spain, May 4–8, 2026)—a major milestone for our work on audio watermarking, provenance, and anti-spoofing security.

1. AURA: A StegaFormer-Based Scalable Deep Audio Watermark with Extreme Robustness

This paper introduces AURA, a scalable deep audio watermarking approach designed for extreme robustness while remaining practical for real-world audio distribution and transformation settings.

"Current deep learning-based audio watermarking are suffering from multiple limitations: insufficient robustness, audible noise, and limited capacity. To address these limitations, we propose Adaptive Universal Robust Audiomark(AURA), an audio watermark framework utilizing our novel Stegaformer module. Stegaformer synergizes a Conformer backbone with Feature-wise Linear Modulation to achieve a deep and robust fusion of the audio and watermark. AURA establishes a new state-of-the-art(SOTA), significantly surpassing existing methods in robustness, particularly against aggressive transformations on diverse audio content, and imperceptible to human ears. Furthermore, this work presents the first scaling law study in audio watermarking, revealing unique architectural principles for effectively scaling model capacity and performance for this challenging task."

2. The Impact of Audio Watermarking on Audio Anti-Spoofing Countermeasures

A first-of-its-kind study showing that audio watermarking can act as a domain shift for spoofing countermeasures and proposing a mitigation framework to preserve detection capability.

3. CompSpoof: A Dataset and Joint Learning Framework for Component-Level Audio Anti-spoofing Countermeasures

This work introduces CompSpoof, targeting a more realistic threat model where only parts of an audio signal (e.g., speech vs. background) are spoofed.

"Component-level audio Spoofing (Comp-Spoof) targets a new form of audio manipulation where only specific components of a signal, such as speech or environmental sound, are forged or substituted while other components remain genuine. Existing anti-spoofing datasets and methods treat an utterance or a segment as entirely bona fide or entirely spoofed, and thus cannot accurately detect component-level spoofing. To address this, we construct a new dataset, CompSpoof, covering multiple combinations of bona fide and spoofed speech and environmental sound. We further propose a separation-enhanced joint learning framework that separates audio components apart and applies anti-spoofing models to each one. Joint learning is employed, preserving information relevant for detection. Extensive experiments demonstrate that our method outperforms the baseline, highlighting the necessity of separate components and the importance of detecting spoofing for each component separately."

Meet Us in Barcelona

We are also proud to announce that OfSpectrum will be an official exhibitor at ICASSP 2026. Beyond the technical sessions where our researchers will present these papers, we invite you to visit our booth in the exhibition area. It’s a fantastic opportunity to see our watermark technology in action, discuss the future of audio provenance, and explore how our security frameworks can integrate into your media workflows. Whether you’re a fellow researcher, a potential partner, or a platform looking to bolster your content integrity, we look forward to meeting you in person and discussing how we can build a more trustworthy audio ecosystem together.

About OfSpectrum

OfSpectrum combines research-grade innovation with production-minded system design, positioning us to deliver a trust layer that scales across creators, platforms, and AI-driven media workflows. Our team is built by research-driven innovators, including Ph.D.-level scientists and active researchers with deep backgrounds in:

• Machine learning and AI security
• Speech and signal processing
• Trust & safety and content integrity technologies

We work hands-on with cutting-edge methods: from modern deep learning architectures to robust signal processing and security evaluation, while staying grounded in what matters for deployment: privacy, copyright protection, data security, and real-world reliability. Our focus is translating frontier research into safeguards that creators and platforms can use at scale.

Interested in partnering? Reach out through our contact page to schedule a meeting during the conference or to learn more about our technology.