# Track me if you can: Ephemeral coin tracing

Crypto.world

> Existing coin-tracing schemes for private payment systems let an authority tag a suspect's funds and follow them through the ledger, but once started a trace can propagate indefinitely through the…

**arXiv**: 2608.09249

**Published**: 2026-08-10

**Authors**: Ignacio Amores-Sesar, Christian Cachin, Rohit Chatterjee, Luiza Soezima, François-Xavier Wicht, Michelle Yeo

## AI summary

## Overview

**Research area:** Applied cryptography and privacy-preserving payment systems — specifically coin tracing for regulated digital money such as central bank digital currencies (CBDCs), institutional stablecoins, and permissioned ledgers.

**Technical level:** Advanced. The paper formalizes a new cryptographic primitive and gives three constructions built from additively homomorphic encryption (exponential ElGamal, Damgård–Jurik, Ring-LWE/BGV), with security games and proofs.

**Scope (one sentence):** The paper introduces *ephemeral coin tracing* (ECT), a primitive in which encrypted tracing tags travel with funds and expire automatically after a publicly fixed number of "hops," so that a law-enforcement trace has a cryptographic limit on its reach rather than a procedural one.

## What This Paper Is About

Existing coin-tracing schemes for private payment systems let an authority tag a suspect's funds and follow them through the ledger, but once started a trace can propagate indefinitely through the transaction graph or persist across all of the target's future transactions. The only thing keeping surveillance targeted and temporary is the restraint of the authority or of a committee holding shares of the tracing key. The paper asks whether tracing reach can instead be bounded by construction — so that even the tracing authority itself cannot extend a trace beyond a public, policy-defined limit.

## Key Contributions

1. **A new primitive, ephemeral coin tracing (ECT).** Every account state carries an encrypted tag recording which traced identifiers its funds hold. When funds move, the sender's tag is degraded and merged with the recipient's. Each tracing contribution expires independently after a policy-defined number of hops and then becomes unrecoverable even to the authority holding the tracing key, without disturbing other live contributions in the same tag.
2. **A formal definition with correctness and security properties.** The scheme is specified as six algorithms — `Setup`, `Tag`, `Verify`, `Degrade`, `Merge`, `Trace` — where `Degrade` and `Merge` require no secret key, so a deployment can assign them to account holders, custodians, validators, or other designated parties. The paper defines two correctness conditions (trace correctness, merge correctness) and four security properties (indistinguishability, ephemeral traceability, unforgeability, unlinkability).
3. **A general "nilpotent" construction template and three instantiations.** ECT is built from additively homomorphic encryption over a plaintext ring containing a publicly applicable nilpotent element ν of index h (ν^h = 0, ν^(h−1) ≠ 0). Degradation multiplies the plaintext by ν; merging adds plaintexts. The instantiations are exponential ElGamal in a group of order q^h, Damgård–Jurik encryption, and Ring-LWE with BGV ciphertexts over R = Z[X,Y]/(X^N0+1, Y^h), the last providing post-quantum security, each with security proofs and a comparison.
4. **Efficient merging of tags at different degradation depths.** Because tags issued at different times sit at different depths, merging must preserve each identifier at its current depth. The paper states that ECT is, to its authors' knowledge, the first coin-tracing scheme to handle such merging efficiently, using a two-layer positional encoding (base-q digits for hop depth, base-r digits for identifier slots within a depth).

## Main Findings

- **Expiry is exact and permanent, even against the authority.** Degradation maps plaintext m to q·m mod q^h; after h applications q^h·m ≡ 0 (mod q^h), and zero stays zero. The security game for ephemeral traceability requires that after k ≥ h′ degradations, `Trace` returns bottom with all but negligible probability. In the exponential ElGamal and Damgård–Jurik instantiations the guarantee is stronger: a fully expired tag is distributed exactly like a dummy tag, even to the holder of the tracing secret key, because degradation rerandomizes.
- **Expiry is per contribution, not per tag.** A single tag can carry several identifiers at different depths, and each expires on its own schedule while the others remain recoverable. The paper's Figure 1 illustrates this with three identifiers arriving at two different depths.
- **Tracing status and remaining budget are hidden from users.** Semantic security hides whether a tag contains zero or an identifier and also hides its current depth, so a user learns neither whether they are being traced nor how much budget remains.
- **Capacity is bounded by public parameters.** A tag can distinguish at most ⌊log_r q⌋ identifier slots; supporting more requires larger public parameters, a change the authority cannot make silently. In the reported evaluation at 128-bit security and h = 10 hops, exponential ElGamal traces 28 identifiers in 1.33 s and adds a few kilobytes per transaction on a commodity laptop. The Damgård–Jurik and Ring-LWE parameters support 341 and thousands of identifier slots respectively, at the cost of larger tags; benchmark timings for those two instantiations are not reported in the supplied text.
- **Exponential ElGamal is recommended for small identifier sets.** The paper states it is typically the best choice for deployments seeking to cap the authority's tracing power at a small set of identifiers, while larger-capacity deployments use the other constructions.
- **Merging is safe under a bounded fan-out.** In the two-state account model each outgoing transfer copies a tag contribution to the sender's change state and the recipient's state, so d = 2 and an identifier's multiplicity is at most 2^(h−1) at any live depth. Choosing r > 2^(h−1) prevents carries between base-r positions and keeps the identifier multiset uniquely decodable.
- **Evasion cost is set by the ledger's hop policy, not by the cryptography.** A hop may be charged on every outgoing transfer, once per epoch or block-height interval, only above a value threshold, or upon interaction with an enrolled counterparty. If every transfer counts, cheap "breadcrumb" payments may exhaust a budget; minimum amounts, rate limits, or counterparty requirements make exhaustion consume time, meaningful value, or identified counterparties instead. Charging a hop only to the sender's tag prevents third parties from exhausting a target's budget via unsolicited incoming payments.
- **Residual trust assumptions remain.** Identifiers are assigned by the authority, so users sharing an identifier slot are indistinguishable to `Trace`, and reassigning a slot before its old contributions expire risks obscuring the trail for up to h hops. The paper also relies on the host ledger to bind tags to states, authenticate fresh entries, prevent reuse, and enforce issuance quotas — limits on which accounts may be tagged are a separate deployment policy, not part of the primitive.

## Methodology in Plain English

The researchers designed a mechanism rather than measuring an existing one. The core trick is to store tracing information inside ordinary encryption in a form that self-destructs through arithmetic: the plaintext space is chosen so that repeatedly applying one public operation (multiplication by a special element ν) eventually maps everything to zero and can never come back. For the main construction this means encrypting with exponential ElGamal in a group whose order is a prime power q^h — a choice that would normally be avoided in cryptography, but here is essential, because multiplication by q in the exponent destroys information after h steps.

Two further ingredients make the scheme usable. First, ciphertexts are additively homomorphic, so multiplying two ElGamal ciphertexts adds their plaintexts: that is how tags merge when funds from several sources combine, with no decryption and no interaction. Second, identifiers are encoded positionally: identifier i is stored as r^(i−1), so a merged plaintext has one base-r digit per identifier recording how many copies are present, and a separate base-q layer records the hop depth. The authority decrypts a tag and reads off both layers to recover which identifiers are still live and how deep each one is. Degradation is applied by whoever builds the payment (with the host ledger's proof binding the update to the hidden input state), and it rerandomizes the ciphertext so tags cannot be linked across steps. Finally, the authors formalize the primitive and prove each construction meets the definitions.

## Why This Matters

**Impact on research.** Prior coin-tracing work bounds *who* may trace — for example by splitting the tracing key among a committee, as in PEReDi and UTT — but not *how much* may be traced; the paper argues that running secure multi-party computation on every transaction to enforce bounds is too costly at ledger scale. ECT reframes tracing reach as a cryptographic parameter fixed at setup, and introduces a template (homomorphic encryption over a ring with a public nilpotent element) that may be reusable for other privacy mechanisms whose authority needs a built-in horizon.

**Real-world applications:**

- **CBDCs and regulated stablecoins**, where privacy must coexist with financial-crime investigation and supervisors want surveillance that is provably limited.
- **Custodial wallets and regulated exchanges**, which can perform degradation as a public, keyless operation without involving the authority.
- **Interbank and cross-institution networks**, where a trace can follow funds across participating banks once off-ledger evidence or a collaborative risk analysis identifies a target.
- **Permissioned/enterprise ledgers** whose account states can carry, update, and merge encrypted tags.

**Industry relevance.** The scheme is deliberately lightweight: degradation needs no secret and no interaction with the authority, tags are a few kilobytes at the reported parameters, and the authority's power is capped by public parameters rather than by policy promises. That combination is what regulators and payment operators would need before deploying private payments at scale — though the paper notes that capacity increases require a public parameter change, so real deployments must size the identifier space in advance.

## Future Directions

- **Choosing and analyzing hop policies.** The primitive leaves open which events trigger a hop; the paper sketches trade-offs between per-transfer charging, value thresholds, rate limits, epoch-based degradation, and counterparty requirements, but a systematic analysis of evasion cost under each policy is left to deployment design.
- **Raising capacity without weakening the bound.** Exponential ElGamal supports only 28 identifier slots at the reported parameters, while Damgård–Jurik and Ring-LWE reach 341 and thousands at the cost of larger tags. Reducing that size overhead — and defining a sound process for public parameter changes — is an open engineering question.
- **Integrating with concrete ledgers.** The paper separates tag operations from the host system, delegating transition enforcement to NIZKs, direct public verification, or trusted-execution attestation, and leaves the design of that enforcement and of issuance quotas or slot-uniqueness policies to future deployment work.
- **Multi-authority and governance questions.** Since identifier slots are assigned by a single authority and users sharing a slot are indistinguishable to `Trace`, how to split issuance across jurisdictions or institutions — and what happens when a slot must be reassigned before its old contributions expire — remains open.

## Target Audience

Cryptographers and security researchers working on privacy-preserving payments, anonymous credentials, and homomorphic-encryption applications; designers of CBDC and regulated stablecoin architectures; and policy or compliance engineers who need to reason precisely about how far a lawful tracing capability can reach. The paper assumes comfort with security games, IND-CPA security, and lattice-based encryption, so readers without a cryptography background will find the formal sections demanding, while the introduction and the deployment discussion in Sections 1 and 7 are accessible to a policy-oriented audience.

## Authors’ abstract

Privacy-preserving payment systems are well understood, yet concerns about their misuse for financial crime have led to only limited adoption in regulated settings such as central bank digital currencies (CBDCs) and institutional stablecoins. Tracing is one tool for addressing these concerns: acting on external evidence implicating a user, law enforcement follows the suspect's funds through the ledger to uncover laundering routes and accomplices. Existing coin-tracing…

[The full abstract on arXiv](https://arxiv.org/abs/2608.09249)

[Read the original paper](https://arxiv.org/abs/2608.09249)

---

Canonical: https://crypto.world/paper/track-me-if-you-can-ephemeral-coin-tracing-2608.09249
