Call for Papers: Workshop on Security Testing and Assurance for LLMs and Agents (STALA) 2027
The Workshop on Security Testing and Assurance for LLMs and Agents (STALA 2027) invites original research on the security of LLMs and AI agents, with a particular emphasis on systematic vulnerability discovery, rigorous defense evaluation, security testing methodologies, and evidence-based assurance.
As LLMs and agents increasingly interact with external tools, sensitive information, and critical infrastructure, their security must be evaluated beyond isolated attack demonstrations. STALA aims to connect practical attack and defense research with systematic testing and assurance methodologies, advancing our understanding of what has been tested, what remains vulnerable, and what security claims can be supported by evidence.
We welcome novel techniques, empirical studies, benchmarks, tools, and experience reports that contribute to the security testing and assurance of LLM-based and agentic systems. Check the workshop website for more details.
Areas of Interest
Topics of interest include, but are not limited to:
Vulnerability Discovery
- Jailbreak and prompt-injection attacks
- Agent hijacking through tools, APIs, and MCP
- Retrieval-augmented generation (RAG), memory, and context poisoning
- Training-data poisoning and backdoors
- Privacy leakage and information extraction
- Multi-agent and delegation attacks
- Insecure code generation and AI-on-hardware attacks
Defense Evaluation
- Adaptive evaluation of guardrails and safety defenses
- Sandboxing, permission management, and access control for agents
- Runtime monitoring and containment
- Security–utility trade-offs
- Robustness of safety alignment
- Backdoor detection and information-flow protection
Security Testing and Coverage
- Security fuzzing and automated red teaming
- Adversarial test generation
- Stateful and long-horizon agent testing
- Differential and metamorphic testing
- Coverage metrics for prompts, behaviors, and agent trajectories
- Neuron-, circuit-, expert-, and routing-level coverage
- Security benchmarks and testbeds
Security Assurance
- Test adequacy and stopping criteria
- Residual-risk estimation and failure-discovery rates
- Confidence and calibration of security evidence
- Probabilistic and compositional assurance
- Integration of testing and formal verification
- Continuous security regression testing
- Quantitative criteria for secure deployment
Submission Guidelines
STALA 2027 welcomes the following paper categories:
- Full Papers: Up to 8 pages, excluding references and appendices.
- Short, Position, and Experience Papers: Up to 4 pages, excluding references and appendices.
We particularly encourage early-stage research, innovative attack techniques, benchmarks, datasets, negative results, and practical experience reports.
All submissions must be written in English and submitted as PDF files using the NDSS 2027 paper formatting guidelines. Papers must use US letter size, a two-column layout, Times font of at least 10 points, and line spacing of at least 11 points.
The submission portal for papers is: https://stala27.hotcrp.com/
The review process is double-blind. Submissions must be anonymized and must not substantially overlap with previously published or simultaneously submitted work.
Each submission will receive two to three independent reviews. Evaluation will emphasize technical quality, clear threat models, rigorous experimental evidence, and relevance to security testing and assurance.
Authors are encouraged to release code, datasets, and other research artifacts.
All submissions must follow the applicable NDSS 2027 policies, including those concerning research ethics and generative AI disclosure.
Publication and Presentation
Accepted papers will be published by the Internet Society in the NDSS Symposium 2027 co-located workshop proceedings before the workshop.
At least one author of each accepted paper must register for the workshop through NDSS and present their work.
Important Dates
- Paper Submission Deadline: 11 December 2026
- Author Notification: 1 February 2027
- Camera-Ready Deadline: 19 February 2027
- Workshop: 22 March 2027
All deadlines are 23:59 Anywhere on Earth (UTC−12).