Every problem here is real.
Set by the authority that owns it. We publish detailed summaries here from launch day, so you can pick with full context when you register. Shortlisted teams receive the complete brief and sample data on 24 September.
AI-driven, risk-based SOC incident triage
Correlate security alerts, assess risk and impact, and prioritise the critical incidents with clear, explainable reasoning, cutting through alert fatigue and false priorities.
Build target
Ingest a sample alert stream, correlate related alerts, risk-score by impact, and output an explainable, prioritised queue.
Adaptive cybersecurity learning and drill platform
Assess a learner's proficiency in a security sub-domain through real interaction, then adapt the training, exercises, and hands-on labs to their level.
Build target
Pick one sub-domain, assess proficiency through interaction, and adapt the next set of exercises to the user.
Automated QBOM framework for post-quantum readiness
Identify, inventory, and track quantum-relevant cryptographic components, assess post-quantum readiness, and guide a prioritised migration.
Build target
Scan a sample codebase or inventory, produce a Quantum Bill of Materials of cryptographic usage, and flag readiness with a prioritised migration list.
Sanchar Saathi force-multiplier
Build a feature, integration, or intelligence layer that measurably improves outcomes on the Sanchar Saathi citizen-protection platform.
Build target
Pick one: an on-device assistant that detects a likely fraud call or SMS in real time and drafts a structured report with evidence, a triage layer that clusters incoming reports into fraud campaigns, or a developer API that checks a number or link against fraud signals before a transaction. Demonstrated end to end on sanitised sample data.
Financial-fraud detection and prevention engine
Over a provided synthetic dataset, detect mule-account behaviour, layering, and known fraud patterns, score the risk, and recommend a prevention action.
Build target
Flag suspected mule accounts and fraud rings with explainable reasons, and output a prioritised case list an investigator or bank could act on.
DPDP compliance and consent-audit engine
Scan a sample application and its data flows, map how personal data is collected, used, and retained, then check it against DPDP principles: consent, purpose limitation, retention, and data-principal rights.
Build target
Ingest a sample app's data map or API surface, detect DPDP gaps, and produce an auditable compliance scorecard with a prioritised, fix-first remediation list.
Shadow AI discovery and governance
Employees are pouring company data into unsanctioned AI tools, browser extensions, desktop LLM apps, and cloud AI APIs faster than security teams can see it happen. Discover unsanctioned AI usage across one surface (endpoint or cloud), assess the risk of each instance, and give a security team a way to govern it.
Build target
Pick one surface: endpoint (browser extensions, web app usage, desktop LLM apps, local models) or cloud (unsanctioned LLM/AI API calls and unapproved AI-enabled SaaS from network or cloud logs). Inventory what you find, risk-score each usage by data sensitivity, sanctioned-vs-unsanctioned status, and compliance impact, and produce a governance action (allow, warn, block, or route to a sanctioned alternative) with reasoning. Demonstrate a discovered inventory, the risk scoring, a seeded data-leak being caught, and the governance recommendation on synthetic telemetry.