Session 1 workshop proposal (chain analysis)

privacy-track/module-01-chain-analysis/WORKSHOP.md

@@ -0,0 +1,132 @@
+This plan replaces the one in README.md
+
+# 0:00–0:10 — Set the investigation
+
+Give everyone access to the same private Signet node.
+
+Present one realistic case:
+
+We know:
+
+- Address A received funds from a ransomware payment.
+- We know the approximate time and amount.
+
+We want to find:
+
+- Where did the funds go?
+- Which addresses appear to belong to the same actor?
+- What behavioral pattern is being used?
+- Where does our evidence become uncertain?
+
+Don't explain what an input/output/UTXO is. Assume all of that.
+
+# 0:10–0:30 — Investigation 1: Fund tracing
+
+Students work directly with bitcoin-cli / RPC.
+
+Give them only the starting address.
+
+Their task:
+
+1. Identify relevant transactions.
+2. Follow the funds.
+3. Record the path.
+4. Identify splits, consolidations and dormant periods.
+5. Produce a small graph of the flow.
+
+5-minute group discussion:
+Different groups will usually choose different interpretations of the same flow. That's useful.
+
+# 0:30–0:50 — Investigation 2: Address attribution
+
+Now give them:
+
+We know:
+- 3 addresses belong to Exchange X.
+
+We want to find:
+- Which other addresses are likely controlled by Exchange X?
+
+Students investigate using evidence such as:
+- common-input relationships
+- consolidation
+- sweep patterns
+- change behavior
+- repeated transaction structures
+- temporal patterns
+
+Crucially, require:
+- Address -> evidence -> confidence, rather than simply "I think these belong together."
+
+This introduces the distinction between clustering and attribution.
+
+# 0:50–1:00 — Break / compare findings
+
+Put 2–3 competing attribution hypotheses on the screen.
+
+Ask:
+> Which evidence actually increases confidence?
+
+This is where I would introduce the idea that on-chain attribution is usually probabilistic rather than cryptographic proof of ownership.
+
+# 1:00–1:30 — Investigation 3: Behavioral analysis
+
+Give them a cluster containing, say, 30–50 addresses.
+
+Don't tell them what the entity is.
+
+We know:
+- These addresses are related to one organization.
+We want to find:
+- What kind of organization/activity does the cluster resemble?
+
+Have them calculate/visualize:
+- transaction frequency
+- active periods
+- holding time
+- fan-in / fan-out
+- transaction-size distributions
+- number of counterparties
+- consolidation behavior
+- periodic sweeps
+- dormant → active transitions
+
+The important part is that **the answer isn't a label**.
+
+They should produce something like:
+
+"This cluster appears to operate as a payment processor because..."
+
+# 1:30–1:50 — Red team / blue team
+
+This is the part I think would make the workshop distinctive.
+
+Give them a deliberately simple detection rule:
+
+> Detector: flag clusters with high fan-out + short holding times + repeated transactions.
+
+Split the room:
+
+Blue team: improve the detector.
+
+Red team: construct synthetic Signet behavior that causes the detector to fail without simply changing the underlying economic story.
+
+Then compare:
+
+- What did the detector catch?
+- What did it miss?
+- Which features were actually useful?
+- Which were easy to manipulate?
+
+This turns chain analysis into an adversarial problem rather than another blockchain tutorial.
+
+# 1:50–2:00 — Final investigation
+
+Give everyone one final question:
+
+> We know: Address A is connected to Entity X.
+> We want to find: What is the strongest evidence that Address B is also controlled by X?
+
+Each group gets 60 seconds to present its conclusion.
+
+Finish by showing the "ground truth" that you built into the Signet scenario.