USENIX Security 2025 Live: Crypto DoS, Secure Graph Analysis, Video Forensics

USENIX Security, one of the top four conferences in information security, will be held in Seattle from August 13–15, 2025. The live paper showcase will focus on cutting‑edge system security research, presenting three breakthrough studies covering cryptographic DoS vulnerabilities, efficient secure graph analysis, and proactive video forensics.

Paper 1: X.509DoS – Exploiting and Detecting Denial‑of‑Service Vulnerabilities in Cryptographic Libraries using Crafted X.509 Certificates

Most prior work on cryptographic vulnerabilities targets confidentiality or integrity, while availability has received little attention. This study systematically investigates DoS‑prone code in cryptographic libraries, discovering that many vulnerable implementations are linked to X.509 certificates. The authors introduce X.509DoS, an attack that uses malformed X.509 certificates to trigger DoS. Using an automated tool for generating malformed certificates and detecting DoS bugs, they found 18 new vulnerabilities across seven mainstream libraries and confirmed 12 known ones, demonstrating the widespread and previously under‑examined threat of X.509DoS.

Paper 2: GraphAce – Secure Two‑Party Graph Analysis Achieving Communication Efficiency

Graph analysis is vital for tasks such as importance ranking, fraud detection, and anti‑money‑laundering, but secure multi‑party computation (MPC) for large graphs suffers from high communication costs, traditionally bounded by Ω(|V|+|E|). GraphAce combines homomorphic encryption with secret sharing and introduces a novel ChaosTable data structure, reducing per‑iteration communication to Θ(|V|) while keeping computation at Θ(|V|+|E|). Evaluations show speedups of up to tens of thousands of times and communication reductions of up to 99.99%, enabling secure analysis of graphs with over one million vertices and 132 million edges.

Paper 3: RollingEvidence – Autoregressive Video Evidence via Rolling Shutter Effect

In response to the surge of deepfakes and video tampering, the authors propose RollingEvidence, which turns the rolling‑shutter artifact of CMOS cameras into a security feature. By embedding a rolling stripe signal into each video frame, the system creates a high‑dimensional, self‑evolving cryptographic trace that is impossible to forge and can be reliably verified. A deep neural network combined with probabilistic models aligns verification signals to detect anomalous frames. The paper validates the approach through theoretical analysis, a prototype implementation, and large‑scale experiments, demonstrating strong protection against video manipulation.

The live session will feature the authors—Shí Bīng (Alibaba Security), Yú Jìpíng (Tsinghua University/Ant Research), and Zhāng Língfēng (Ant Digital Technology)—who will share design insights and validation results.