Soteris Demetriou
Scholar

Soteris Demetriou

Google Scholar ID: 4utItXUAAAAJ
Associate Professor, Imperial College London
Computer SystemsMobile ComputingCybersecurityPrivacyApplied Machine Learning
Citations & Impact
All-time
Citations
1,728
 
H-index
17
 
i10-index
21
 
Publications
20
 
Co-authors
74
list available
Resume (English only)
Academic Achievements
  • Paper 'Leaky Diffusion: Attribute Leakage in Text-Guided Image Generation' accepted to PETS 2025 (May 2025)
  • Paper 'DiDOTS: Knowledge Distillation from Large-Language-Models for Dementia Obfuscation in Transcribed Speech' accepted to PETS 2025 (October 2024)
  • Paper 'ClaritySpeech: Dementia Obfuscation in Speech' accepted to Interspeech ’25 (May 2025)
  • Paper 'Understanding Dementia Speech Alignment with Diffusion-Based Image Generation' accepted to Interspeech ’25 (May 2025)
  • Paper 'Adversarial 3D Virtual Patches using Integrated Gradients' accepted to SafeThings '24 (March 2024)
  • Paper 'Adversarial 3D Virtual Patches using Integrated Gradients' won Best Paper Award at SafeThings '24 (May 23, 2024)
  • Paper 'MobileConfig: Remote Configuration Management for Mobile Apps at Hyperscale' accepted to NSDI '24 (December 2023)
  • Served on the Technical Program Committee of PETS 2025 (May 2024)
  • Served on the Technical Program Committee for the Hardware/Cyber Security and Privacy Track of IEEE COINS 2024 (March 2024)
  • Served on the Technical Program Committee of ACM CCS 2024 (December 2023)
  • Served on the Technical Program Committee of ACM ASIA CCS 2024 (December 2023)
Background
  • Associate Professor of Systems Security at Imperial College London
  • Director of Imperial’s Academic Centre of Excellence in Cyber Security Research (ACE-CSR)
  • Leads the Applications, Platforms, and Systems Security (APSS) research group
  • Research focuses on security and privacy issues in computer systems and machine learning models
  • Work includes uncovering novel side-channel and confused deputy threats in Android OS, privacy risks from mobile advertising, IoT vulnerabilities in smart homes, privacy and safety issues in traditional and generative AI models, and vulnerabilities in 3D object detectors used in connected and autonomous vehicles
  • Has designed access control schemes, anomaly detection, trusted computing, and privacy-preserving mechanisms to strengthen mobile and CPS security
  • Collaborated with Meta on large-scale distributed computing systems related to function-as-a-service, microservice routing, continuous deployment, and configuration systems