SAKSI

Inspiration

In both 2024 and 2026, Indonesian medical residents have succumbed to death due to months of harassment and extortion, which were allegedly unknown to the medical program's administration. The official Ministry of Health hotline has recorded 2,621 reports; in 2020, it was revealed that 63% of lecturers, who were aware of sexual violence happening in campus, did not report the incident. All the 4,303 campuses of Indonesia are legally obligated to establish a reporting task force – none of which has proof of any report as the institution maintains the database.

What it does

SAKSI ("Witness") distinguishes the thing that was reported from the act of reporting it. Any student or resident creates a report through any web browser without creating an account and without any wallet. The report is encrypted in their own device; only its cryptographic hash is recorded in a public L2 blockchain with a timestamped block. The reporter gets a Sealed Receipt which can be shown to the relevant ministry, to the court or to the journalist independently of the institution. The task force within the institution should confirm and update the report through on-chain actions using their wallets; response time becomes transparent, trackable information. A report ignored for 7 days can be submitted to the ministry with one mouse click. An open weight AI model (with Featherless.ai) creates structured case files from narrative text reports and anonymized summaries for the task force.

How we built it

Next.js + WebCrypto (AES-256-GCM, client-side) · Solidity registry contract using Lisk Sepolia blockchain via Foundry · gas-powered relayer + event listener (viem) · Supabase for storing ciphertext only · Featherless.ai API that is OpenAI-compatible for intake format and summary generation · deployed on Vercel and testnet. Four Features: Sealed Report Submission, Task-Force Console, Public Accountability Ledger, Verify & Escalate.

Challenges we ran into

• Real, current, openable data. Numbers on bullying on campus and in hospitals in Indonesia are found in ministry press conferences, news stories and paywalled sources, and vary per source – 2,621 cases reported through a hotline from the Health Ministry versus dozens of complaints directly received by the human rights committee. I quoted only figures which I was able to open and trace back to an identified authority or document.
• Addressing "why not a database?" truthfully. My initial draft had too much on chain. Final draft kept only the hash, timestamp and actions of the task force while keeping the content encrypted off-chain and secured by a key held by a reporter. That answered the blockchain dilemma as well as the data security one.
• Anonymous vs. eligible. Verifying "I am an eligible member of this organization" without disclosing my identity is a difficult challenge. For the MVP version in 24 hours, I decided on one-time use codes issued by the organization and zero-knowledge proof of membership for later versions instead.
• Scoping down to 24 hours. Two-way anonymous messaging, mobile app, and AI hosted locally were all removed. Four features that have full end-to-end demonstration via a public URL trumped eight that do not.

Accomplishments that we're proud of

• A problem statement with each statistic having its source, including incidents that occurred in January and July 2026.
• A solution where the organization that reports this data cannot change, back-date, or dispute the data in any way, but does not reveal any personally identifiable information to the blockchain.
• Seven-day escalation requirement: non-compliance becomes publicly available data that is enforceable using the time lock on the contract.
• A construction plan for each hour to build the solution, such that the proposal is a plan that can be executed at the end.

What we learned

• The issue of under-reporting is one of trust, not awareness: 63% of lecturers aware of the incidence of sexual violence on their campus did not report it; the operator of the reporting channel is more important than its design.
• The portion of the blockchain that we can use in this case is extremely small: about 100 bytes for each report. Its value does not lie in the stored information itself, but in the fact that no one can edit it.
• We have regulation for our market segment: every Indonesian campus must have a task force for reporting issues, and cannot prove it received any reports. This gap SAKSI closes.
• Read the rubric first and only then write. "Feasible to build in 24 hours" counts 30% of your grade, so it is a scoring decision, not a compromise.

What's next for SAKSI

Use zero-knowledge membership proofs instead of one-time access codes, anonymous bidirectional communication between the reporter and the task force, on-premises model deployment for the ministries, and integration with the current Kemenkes reporting system.

Built With

• featherless
• foundry
• lisk
• nextjs
• postgresql
• solidity
• supabase
• typescript
• viem
• webcrypto