🔬 Research Presented at KDD25: Graph Reconstruction Attacks on Secure Protocols

We’re excited to share that our research on graph reconstruction attacks was presented at KDD25 (Knowledge Discovery and Data Mining 2025), one of the premier conferences in machine learning and data science.

Conference Experience

KDD25 was an incredible experience, bringing together researchers from around the globe to discover the latest advancements in machine learning. The conference showcased outstanding work spanning from core AI foundations to diverse real-world applications, providing valuable insights into the current state and future directions of the field.

Research Presentation: Graph Reconstruction Attacks

Our presentation focused on graph reconstruction attacks, a critical security vulnerability that demonstrates how even secure protocols can inadvertently leak information about their inputs. This research highlights the importance of thorough security analysis in privacy-preserving technologies and secure computation protocols.

Key Research Contributions

  • Security Vulnerability Discovery: Identified information leakage in supposedly secure protocols
  • Attack Methodology: Developed techniques for reconstructing graph structures from protocol outputs
  • Privacy Implications: Demonstrated how partial information can lead to complete graph reconstruction
  • Defense Strategies: Proposed countermeasures to prevent information leakage

Responsible AI Day

We were honored to present our work during the Responsible AI Day track, which emphasizes the importance of developing AI systems that are not only powerful but also secure, ethical, and privacy-preserving. Our research directly contributes to this mission by identifying potential security weaknesses in AI and ML systems.

Research Impact

This work has significant implications for:

  • Privacy-Preserving Technologies: Ensuring truly secure computation protocols
  • Graph Analysis Systems: Protecting sensitive network structures
  • AI Security: Building robust and trustworthy AI systems
  • Industry Applications: Securing enterprise graph analytics platforms

Collaboration & Recognition

A big thank you to the KDD25 organizers for hosting such an exceptional conference, and to our brilliant coauthors for making this research possible. This collaboration demonstrates the power of combining academic research with practical security analysis.

This research presentation at KDD25 represents our commitment to advancing both the theoretical foundations and practical security of privacy-preserving technologies. By identifying vulnerabilities in secure protocols, we contribute to building more robust and trustworthy systems for real-world applications.

For more details on our research, check out the full paper: GRAND : Graph Reconstruction from potential partial Adjacency and Neighborhood Data