Train cybersecurity QA models with paired certification questions and expert explanations focused on incident response.

Train models to analyze iOS application security evidence and deliver expert-level answers grounded in platform knowledge.

This dataset is used to identify whether workers are wearing helmets and if they are wearing them correctly, supporting intelligent monitoring and safety management in the security field.

The Elderly Fall Detection Dataset focuses on monitoring elderly falls in the home environment to enhance family safety.

Verify whether microsegmentation policies are correctly reconciled and enforced, even when controller execution encounters complex failure states.

Trace kernel code paths and verify privilege boundaries with source samples designed for coding agent analysis.

Help models understand cross-packet IDS/IPS session logic and detect implementation issues through structured verification of source code and state rules.

This dataset is used for recognizing the action of people climbing over fences, applicable to various security scenarios.

Evaluate whether code enforces continuous zero-trust access control using session lifecycle signals and supporting code evidence.

This dataset is used for detecting the wearing status of safety equipment for personnel within construction areas.

Match cyber incident claims to case-file evidence with labels for sufficiency, supporting passages, and evidentiary gaps.

Explore Zero Trust service failure scenarios and uncover exception-handling defects that could undermine access control.

Train models to identify new disaster-related changes more reliably using pre- and post-disaster mapping records and explicit inference rules.

The cleanroom personnel dress code compliance detection dataset is used to identify and improve personnel dress compliance in cleanroom environments.

Combine diverse rescue records with failure analyses to help models produce evidence-based, professional incident reviews.

Train models to distinguish authentic botnet advisories, reposts, impersonation, and inconclusive claims by comparing pages with publisher references.

Train models to infer building damage categories from UAS survey observations and explicit rules, while recognizing when evidence is insufficient.

This dataset is used to identify suspicious repeated glancing actions in surveillance scenes to enhance security monitoring effectiveness.

Explore source code and structured annotations for identity context propagation failures in zero trust systems, and verify the integrity and trustworthiness of information used in access decisions.

This dataset is used to detect reverse movement on escalators, prolonged stay, and sitting on escalators, applicable to locations like subways and shopping malls.