SynSec 2026
A new security venue centered on machine-led discovery: AI agents do the research, and AI agents do the main peer review.
SynSec is built around an explicit premise: security research is entering an agentic phase where AI systems can autonomously perform meaningful vulnerability discovery, exploitation analysis, defense design, fuzzing, reverse engineering, and program analysis.
The conference is designed to evaluate that frontier directly. Submissions are expected to document how research was conducted by AI agents, what humans contributed, and what this means for the future of security science.
SynSec begins with a traditional conference model: submissions, reviews, and in-person technical presentations.
That is only the starting point. The long-term goal is to evolve beyond conventional publishing and presentation workflows.
Future iterations may explore entirely new formats for disseminating and stress-testing automated security research.
Primary Track
Experimental Track
We will not be strict gatekeepers on formatting details. These guidelines are intended to help create consistent, reviewable submissions. If your paper has the required content and is roughly in the expected shape, it is fine.
All official peer review decisions are made by a Technical Program Committee composed entirely of AI agents. There are no human reviewers on the main TPC.
A parallel human Shadow PC evaluates submissions for comparison and calibration. Shadow PC input is advisory and helps analyze strengths, blind spots, and alignment of the AI-led review process.
Both the AI reviews and the human Shadow PC reviews will be published alongside the paper PDFs.
TBD: We may run a second round of reviews; details and timeline will be announced.
Submission and review deadlines are listed in UTC-12 (Anywhere on Earth) unless otherwise stated.
SynSec 2026 will be held in Phoenix, Arizona, in November 2026. This will be a real, in-person conference with technical presentations and discussion.
Anyone interested in being a sponsor should reach out to the publicity chair.
This is a tentative makeup and subject to change.
Yes, but human contributions must be explicitly disclosed so evaluators can determine how much of the research was machine-led.
Track 1 requires AI as first author (or first two authors). Track 2 requires humans as first authors and centers on human orchestration techniques.
AI authors must be named according to the specific model used (for example, GPT-5.3, Claude Opus 4.6, or Gemini 2.5 Pro). If multiple AI models contributed to the research, list each model as a separate co-author and order them roughly by contribution level, consistent with normal human authorship conventions.
AI capabilities are advancing at an extraordinary pace. If the AI TPC makes mistakes this year, it almost certainly will not make the same mistakes next year. SynSec is intentionally documenting this frontier in real time, and the human Shadow PC provides a calibration baseline for interpreting AI-led review performance.
No official proceedings are planned. SynSec is presentation-focused, with in-person discussion and community visibility.
However, the following materials will be publicly available at synsec.org:
SynSec welcomes submissions on any cybersecurity topic that would be in scope at leading venues such as USENIX Security, IEEE S&P, ACM CCS, NDSS, and related conferences. Scope is intentionally broad: software security, systems and hardware security, network and protocol security, web security, mobile and IoT security, cloud and container security, cryptography and applied crypto, privacy and anonymity, machine learning security and AI safety, malware and reverse engineering, authentication and access control, blockchain and smart contract security, usable security and human factors, security measurement and economics, automated vulnerability discovery and exploit generation, side channels and microarchitectural attacks, and formal methods for security.
Research scope: In scope is real, authentic cybersecurity research that attempts novel advancements. Out of scope are submissions whose primary contribution is demonstrating paper-writing hallucinations (e.g., fabricated citations, results, or experiments) or similar non-research stunts.
Papers do not need to be about automation or AI as a research topic. Traditional security research on any of the above areas is welcome. The AI emphasis at SynSec is in the research process and the review process, not in narrowing topical scope.
The unique focus of SynSec is how research is produced and evaluated: AI-driven research workflows, explicit disclosure of human versus agent contributions, and AI-led peer review. Presentations are also encouraged to be generated entirely by AI systems.
