OPTIMIST PhD Forum
The OPTIMIST PhD Forum is oriented at Ph.D. students working on topics related to AI and side-channel analysis (SCA). The goal is to create an interactive environment where students:
- Present short research pitches,
- Share their recent or published work,
- Build a strong network and exchange ideas.
Interested in joining the next session?
Session 1: 16 April 2026
Opening Talk
- Trevor Yap: Breaking the blindfold: Deep learning-based blind side-channel analysis
Blind side-channel analysis (SCA) aims to recover secret keys without access to plaintext or ciphertext, a setting where existing approaches have shown limited practical success. In this work, we introduce Deep Learning-based Blind Side-channel Analysis (DL-BSCA), a framework that leverages deep neural networks for blind key recovery. We also propose Multi-point Cluster-based (MC) labeling, a method that improves trace labeling by exploiting dependencies between leakage variables across multiple sample points. We evaluate our approach on four real-world datasets spanning AES, ASCON, and Kyber implementations on both AVR XMEGA and ARM STM32F4 platforms. Our results demonstrate successful blind key recovery, including the first blind SCA attack against desynchronization countermeasures.
Discussion
The first session of the OPTIMIST PhD Hour focused on introducing attendees to one another and fostering collaboration among researchers working in AI and side-channel analysis. Participants shared their academic backgrounds, discussed the research problems they are currently working on, and engaged in networking and brainstorming discussions aimed at finding potential solutions and new research directions. The meeting featured a diverse range of technical discussions and encouraged interdisciplinary collaboration among participants. These topics were discussed during the session:
- Fully blind side-channel analysis in the presence of masking countermeasures.
- Explainability in deep learning-based side-channel analysis.
- Plateau effects in machine learning models used for side-channel analysis.
- Transformers and their applications in side-channel analysis.
Session 2: 30 April 2026
Talk 1
- Arna Roy: A Framework for Scalable Pre-Silicon Side-Channel Analysis
We present a hierarchical and module-centric framework for scalable pre-silicon side-channel leakage assessment of Post-Quantum Cryptography hardware. By isolating security-relevant modules and signals, the framework significantly reduces simulation complexity while enabling efficient and interpretable leakage analysis. Evaluation on ML-DSA-87 demonstrated the ability to detect leakage in masked implementations while achieving substantial efficiency improvements compared to full-design RTL and gate-level evaluation.
Talk 2
- Fatemeh Khojasteh Dana: GlitchSnipe: Toward Localized Voltage Fault Attacks
Voltage glitching attacks were analyzed through the perspective of electromagnetic energy propagation within the power delivery network (PDN). The study showed that different frequency components travel across the chip in distinct patterns, enabling localized fault injection through sinusoidal voltage modulation. Experimental results on FPGA platforms demonstrated that different chip regions exhibit varying sensitivities to injected frequencies, and that even small placement changes can significantly impact resilience against voltage-based fault attacks.
Discussion
During the second OPTIMIST PhD Hour, Dr. Andrew Adiletta introduced the MITRE Side-Channel Lab, an open-source educational platform designed for understanding side-channel analysis and exploring practical applications. The discussion included a walkthrough of one of the interactive notebooks that demonstrates and teaches core SCA concepts through hands-on examples. Participants were also provided access to the online notebook environment through MITRE Side-Channel Lab.