betaFeedback

WorksComputer Science

Baryonic Keys and the Inheritance of Crypto-Assets: Revisiting the Tales From the Crypt Protocol

Abstract

The inheritance of crypto-assets faces a structural obstacle: ownership is the knowledge of a key, so the key either disappears with its owner or has to be shared beforehand. The Tales From the Crypt Protocol (TFCP) addressed one part of the problem, the distributed and privacy-preserving acknowledgment of a death by a blockchain. We revisit it with a new primitive. A baryonic key is a physical object from which bits are obtained only through a measurement, are never stored digitally, and which exists in exactly nn copies, where nn is neither zero nor unbounded. We define the primitive, outline a candidate instantiation based on a replicated micro-fractured relief, and state its limits: digitising a copy is slow and tamper-evident, not impossible. We also outline other uses of the primitive, from post-quantum pre-shared keys to common coins for Byzantine agreement. We then give a threat model for TFCP, review the attack models of the literature that apply to it, and show where it resists and where it does not. We propose TFCP-B, in which decrypting the will requires both a secret shared among registrars and the response of a baryonic key, and in which heirs claim their share through hash commitments, without any prior key or wallet. We detail each change from TFCP to TFCP-B, with its rationale and its price. TFCP-B removes the anonymity loss caused by colluding registrars and blocks several false-alarm attacks. It does not remove the shallow-account attack, remains exposed to transaction censorship during the deliberation period, and introduces new assumptions on custodians and reading devices. The security arguments are informal and the physical figures are design estimates.

Introduction

A crypto-asset is controlled by whoever knows the corresponding private key. This is the source of its resistance to seizure and censorship, and also the source of a difficulty that is seldom addressed: when the owner dies, either the knowledge of the key dies as well, or the key has been given to somebody else in advance, in which case ownership was shared during the owner’s lifetime [32, 33].

Two earlier papers explored this difficulty. The first one [32] proposed the Tales From the Crypt Protocol (TFCP), by which a blockchain can acknowledge the death of an account owner without a trusted oracle and without revealing, before the death, which account belongs to whom. It deliberately left aside the delivery of credentials to the heirs. The second one [33] listed open questions, and suggested in its Section 4.2 that secret shares could be “objectified”: turned into physical objects cloned by an analogue process, from which shared bits would later be obtained by measurements.

The observation behind that suggestion is that digital information admits only two regimes of copying. A secret known to a single person has no copy and disappears with that person. A secret written once in a computer memory can be copied perfectly, silently and without bound. Inheritance requires an intermediate regime: a secret that exists in exactly nn places, with nn small and fixed, and that cannot be duplicated in a second.

Contributions.

  1. We define baryonic keys, physical keys that are clonable a bounded number of times, and outline a candidate instantiation (§ 3). We position the primitive with respect to physical unclonable functions and to recent work on twin and distributable physical functions (§ 2), and outline its uses beyond inheritance (§ 3.4).

  1. We give a system and threat model for death-acknowledgment protocols (§ 4).

  1. We specify TFCP-B, a revision of TFCP built on baryonic keys (§ 5). The will is encrypted under a key derived from two halves that are useless in isolation; heirs claim slots defined by hash commitments. We then make explicit what changes with respect to TFCP, why, and at what price (§ 6).

  1. We analyse TFCP and TFCP-B against attack models taken from the literature on bribery, censorship, front-running, flash loans, fuzzy extractors and physical unclonable functions (§ 7), and state for each of them whether the protocols resist.

  1. We revisit the open questions of [32, 33] (§ 8), compare costs with those of existing services (§ 9) and list the limitations of the approach (§ 10).

Scope. The security arguments of this paper are informal. The physical figures are design estimates that have not been validated experimentally. Both points are discussed in § 10.

Background and related work

The Tales From the Crypt Protocol

We keep the vocabulary of [32]. The Donor is the account of the person whose death has to be acknowledged. The Security Deposit is an account controlled by the Donor and used to signal the death. Witnesses testify to the death by transferring an Ante to the Security Deposit. Registrars stake a Bail and hold Shares of a key that encrypts the Wills, a document linking the Donor’s account to a public Announcement that bears the civil identity of the Donor.

The Donor recruits Registrars, hands them the Shares, and publishes the encrypted Wills and the Announcement. After the death, Witnesses transfer their Antes. Once a Threshold is reached, a Deliberation Time starts. If the Donor’s account moves during that time, the Donor is alive: the process stops and the Security Deposit, which now holds the Antes, becomes spendable by the Donor. Otherwise the Registrars decrypt the Wills, the Acknowledgment is enacted by the network, fees are paid to Registrars and Witnesses, and inheritance transactions are executed. The content of these transactions is outside the scope of [32].

The stated properties are a distributed oracle, the anonymity of the Donor’s account before the Acknowledgment under an honest-but-curious assumption on Registrars, and independence from the underlying blockchain. Two attacks are discussed in [32]: a Donor who hands the key of an almost empty account to an accomplice in order to capture the Antes, and a wealthy Witness who reaches the Threshold alone.

Inheritance of crypto-assets in practice

Deployed solutions have converged on the dead man’s switch, which [33] already criticised for the constant vigilance it requires. Liana adds to the owner’s key a recovery key that only becomes valid after a period of inactivity, the delay being enforced by the Bitcoin network itself [46]; Nunchuk launched an inheritance scheme based on the same on-chain mechanism at the end of 2025 [Spa26]. Collaborative custody services such as Casa rely on multi-signature wallets in which the provider holds one key [5]. Sarcophagus is a dead man’s switch operated by staked third parties [39]. A legal analysis of crypto-asset succession in civil-law systems is given in [6].

These solutions share two limits. First, the trigger is the owner’s silence rather than the owner’s death, so that a short delay may open the succession of a person who is travelling or in hospital, while a long delay postpones every legitimate succession. Second, the heir has to hold a key or a device as of today. Neither the distributed acknowledgment of a death nor the transmission of rights to a party that is not yet known is addressed.

Physical unclonable functions and shared physical keys

Physical one-way functions were introduced by Pappu et al. [34], who used the speckle pattern of a disordered optical token as an identifier that is cheap to make and very hard to duplicate. The large family of physical unclonable functions (PUFs) that followed is surveyed in [26]. A PUF is unique by design: the party that holds it must extract a key and transmit it by conventional means if another party is to share it.

Two recent lines of work depart from uniqueness. Zhong et al. [ZLX+22] fabricate twin PUFs on the same aligned carbon nanotube array; the two devices agree on about 95 % of their bits and are used as a shared key that is never extracted or stored. Luescher et al. [25] propose chemical unclonable functions based on random DNA pools, which are explicitly designed to be distributable: a pool can be split into identical pools held by several parties. A holder who knows the structure of the pool can amplify it further, and the authors add a chemical modification to disable copying.

Baryonic keys belong to this second family. They differ in the property that is taken as primary: the number of copies is fixed once and for all by the destruction of the replication master, and the design goal is the transmission of secrets across decades rather than authentication or communication.

Reconciling noisy physical readings into identical keys is the purpose of fuzzy extractors [15]. Two refinements matter here: reusability, since several enrolments of the same source may leak more than a single one [4], and robustness against an adversary who modifies the public helper data [13]. Modelling attacks, in which a machine-learning model trained on observed challenge–response pairs predicts new responses, have broken several electrical PUF designs [37].

Baryonic keys

Definition

Definition 1 (Baryonic key). A baryonic key with parameters (n,ℓ,τ)(n,\ell,\tau) consists of nn physical objects, the copies, a challenge space C\mathcal{C}, and a public reading procedure which, on a copy and a challenge c∈Cc \in\mathcal{C}, returns a noisy string w(c)w(c). Together with a fuzzy extractor (Gen, Rep), it defines a response R(c)∈{0,1}ℓR(c) \in\{0,1\}^{\ell} and public helper data P(c)P(c) such that the following properties hold.

B1 – Agreement.

For every challenge cc and every copy, Rep applied to a reading of that copy and to $P(c)

returns R(c)R(c), except with negligible probability.

B2 – No digital storage. No representation of the responses exists outside the matter of the copies; P(c)P(c) reveals a negligible amount of information on R(c)R(c).

B3 – Bounded clonability. Exactly nn copies exist. Producing a further copy, or a digital model that predicts responses to arbitrary challenges, requires physical access to a copy for a time at least τ\tau and leaves physical evidence.

B4 – Independence. Knowledge of the responses to a set of challenges gives a negligible advantage in predicting the response to a challenge whose reading does not overlap theirs.

B5 – Durability. The copies are passive and remain readable for decades.

A PUF corresponds to the case n=1n = 1 with τ\tau ideally infinite. The twin PUFs of [ZLX+22] correspond to n=2n = 2. The pools of [25] are distributable, with a number of copies that the holders themselves can increase unless copying is chemically disabled. A baryonic key fixes nn at fabrication and accepts a finite τ\tau.

On the name. The word baryon was introduced by Pais in 1953 [30], from the Greek barýs, heavy, for the family of heavy particles to which the proton and the neutron belong. By extension, baryonic matter is the ordinary matter of which atoms, and therefore all everyday objects, are made; cosmologists use the term to set it apart from what is not made of atoms. We borrow the adjective for a similar contrast. A digital key is information: it has no weight and no location, and it is duplicated at no cost. A baryonic key is a piece of matter: it weighs something, it is somewhere, and its bits do not exist as information until they are measured.

A candidate instantiation

We outline the instantiation that we are studying. All figures in this subsection are design estimates.

Object. A glass master is roughened by micro-fracturing (fine abrasive blasting followed by a mild etch), which produces a random relief with a roughness of 50–200 nm and a correlation length of about 1 μm1\,\mu\mathrm{m}. The relief is replicated nn times by direct nanoimprint [8] onto glass discs of 120 mm in diameter. Each copy is coated with a mirror and sealed under a thin transparent layer. The master is then destroyed. Replication is physical and parallel, so that no file describing the relief exists at any time.

Reading. Optical pickup units designed for optical discs measure a focus-error signal with nanometre sensitivity, and have been used as low-cost profilometers and as sensing heads for atomic force microscopes [21, 38]. The reader uses such a unit to measure the relief at a few sites of 10×10 μm10 \times10\,\mu\mathrm{m}. A challenge is expanded by a hash function into a list of sites; the mapping is public.

Agreement. Two copies read by two devices are expected to differ on a few per cent of the raw bits. Gen is run once, on one copy, and outputs R(c)R(c) and P(c)P(c); Rep is then run on any copy. Each site is used by a single run of Gen, which avoids the reusability issue of [4].

Valid measurement sites per side≈8×107\approx8 \times10^{7}
Net bits per site, at 2% raw error≈7.4\approx7.4
256-bit response, 2% raw error50 sites, read in ≈18\approx18 s
256-bit response, 8% raw error (ageing margin)122 sites, read in ≈35\approx35 s
Number of copies nn2 to 5
Digitisation of one side, τ\tau1–10 h with laboratory equipment; 10–100 h with portable equipment
Fabrication of a copy from a digitised reliefweeks (greyscale nanolithography)

caption: tbl:1: Design estimates for the micro-fractured disc. None of these figures has been validated experimentally.

:::

Status of the properties. B1 follows from the fuzzy extractor once the raw error rate is confirmed. B2 holds by construction for the copies, and for the workshop only if the master is destroyed without having been measured. B4 rests on the absence of long-range correlation in the relief; unlike the delay-based electrical PUFs broken in [37], a response here is a function of local heights only, but this has to be established by measurement. B5 is plausible for a mineral, sealed object and has to be confirmed by accelerated ageing.

B3 holds in a weak form only. A legitimate reading touches a fraction of about 10−610^{-6} of the surface, whereas predicting arbitrary responses requires digitising all of it; but τ\tau is counted in hours, not in years, against an equipped adversary who takes a copy away. The protocol of Section 5 is therefore designed so that a fully digitised copy gives nothing by itself.

The wish list of the second paper

Section 4.2 of [33] listed the properties expected from objectified shares. Table 2 compares them with the candidate instantiation.

Wish in [33], Section 4.2Candidate instantiationRemaining gap
Clones produced without any filedirect nanoimprint from a master that is destroyed afterwardsthe workshop handles the master
Measurement not fixed at production timethe challenge is chosen later, among millions of independent onesthe reader must be specified and archived
Identical measurements on every copyexact agreement through a fuzzy extractorraw error rate to be measured
Scanning should not suffice to produce a clonedigitisation takes hours and leaves tracesscanning suffices to compute responses
Stability over long periodsmineral, passive, sealed objectageing to be measured; readers are perishable
Access to the object should not be enoughobject and challenge are kept apart–

Table 2. The wish list of [33] and the candidate instantiation.

Other uses of baryonic keys

Inheritance is the use that motivated this work, but properties B1–B5 are of interest wherever a small group of parties must share secrets that should exist in no memory. We outline six such uses and relate them to recent work. They are not analysed in this paper, and the caveats of Section 3.2 apply to all of them. In each case the workshop plays the role that a trusted dealer plays in the cryptographic literature, with the difference that it hands out objects rather than digital secrets.

UseWhat the key providesRelated work
Post-quantum pre-shared keys between sitesthe same fresh key at nn sites for each epoch, stored nowhere; a compromised host extracts key material no faster than the reading device allows[17, 41]
Common coin for asynchronous Byzantine agreementidentical random bits at all holders; with one key per subset of parties, a threshold coin without distributed key generation[36, 9, 23, 11, 18, 7]
Randomness beaconresponses committed by hash at setup and revealed on schedule; values are fixed at fabrication and cannot be biased[10, 14]
Threshold signaturesper-subset secrets for pseudorandom secret sharing, re-derived at each use; deterministic nonces[22, 7]
Secure aggregationpairwise masks without a key-agreement phase[2, 29, 24]
Trusted components for replication with 2f+12f + 1 replicasthe secret shared by the trusted components, re-derived at start-up and never stored[44, 20, 1]

Table 3. Uses of baryonic keys beyond inheritance.

Pre-shared keys. RFC 8784 mixes a preshared key into the key derivation of IKEv2 so that traffic recorded today cannot be decrypted later by a quantum computer [17]. RFC 9867, published in November 2025, extends the mechanism to the initial exchange and to rekeying, the earlier specification having assumed static keys [41]. Both leave open how the preshared key is distributed and where it is kept. A baryonic key shared by two or more sites answers both questions: the sites read the same response to a fresh public challenge at each epoch, and the key exists in memory only while it is used. Since responses are independent (B4), a host that was compromised for a limited time has obtained the keys it was able to read during that time and no others.

Common coins. Asynchronous Byzantine agreement needs randomness, classically in the form of a common coin [36, 9]. The coin of [9] relies on a private setup performed by a trusted dealer, and removing this setup is an active line of research: asynchronous distributed key generation [23], consensus with neither trusted setup nor public-key cryptography [11], and a common coin that only requires a silent public setup [18]. A baryonic key takes the opposite route and keeps a dealer, but a physical one. If one key is fabricated for each subset of N−fN - f parties among NN, which is the replicated secret sharing of [7], any coalition of ff holders lacks one key; the exclusive-or of the responses of all keys to a given challenge is then a coin that such a coalition cannot predict before an honest holder reveals its values. One round of messages is needed, since no party holds every key: each holder announces its responses, and a response is accepted once f+1f + 1 holders of the same key agree on it. No key-generation protocol is run, and no computational assumption is added to the physical ones. For N=4N = 4 and f=1f = 1 this requires four keys of three copies each.

Randomness beacons. Distributed randomness beacons are expected to be unbiased and unpredictable [10]; deployed beacons such as drand [14] rely on threshold signatures. With a baryonic key, the Donor-side technique of § 5 applies directly: commitments to future responses are published at setup, and one response is revealed per period and checked by hashing. The values are fixed at fabrication, so no participant can influence them, and every key has several holders, so none can withhold a value. The capacity of a key bounds the frequency of such a beacon.

Threshold signatures. Arctic [22] is a deterministic and stateless threshold Schnorr signature for small groups, built on a verifiable pseudorandom secret sharing whose secret keys are indexed by subsets of signers, in the manner of [7]. Secrets indexed by subsets are exactly what a family of baryonic keys provides: they can be re-derived from the discs at each signature instead of being stored, and they are distributed by the fabrication itself rather than by an interactive ceremony.

Secure aggregation. In the protocol of Bonawitz et al. [2], each pair of clients agrees on masks that cancel in the sum, so that the server learns the aggregate only. Later systems reduce the cost of the setup that establishes these pairwise secrets, by performing it once for many rounds [29] or by letting each client speak once [24]. For a small consortium, baryonic keys held in pairs supply the masks with no key-agreement phase at all.

Trusted components. Byzantine fault-tolerant replication needs only 2f+12f + 1 replicas when each replica hosts a trusted component that prevents equivocation. In one variant of MinBFT [44], the trusted components authenticate their certificates with a secret key that all of them share. Whether trusted components are best used to reduce the number of replicas is debated [20, 1]; in either case a baryonic key in 2f+12f + 1 copies is a way to provision the shared secret without a key ceremony.

System and threat model

Parties

Besides the Donor, the Witnesses and the Registrars of TFCP, we consider: the Custodians, who each keep one sealed copy of a baryonic key; the envelope keepers (typically a notary), who keep sealed envelopes containing challenges; the heirs; the validators of the blockchain; the workshop that fabricates the key; and the reading devices.

Assumptions

Assumption 1 (Cryptography). H is a hash function with 256-bit outputs modelled as a random oracle; KDF is a key-derivation function with the same idealisation; AE is an authenticated encryption scheme.

Assumption 2 (Blockchain). The blockchain is safe and live. A transaction submitted with an adequate fee is included within a bounded delay, unless validators are paid to censor it; the cost of censorship grows with its duration [45, 43].

Assumption 3 (Physical key). The succession key satisfies B1–B5 with the parameters of Table 1.

Assumption 4 (Separation). No party other than the Donor holds both a copy of the key and an envelope before the Acknowledgment.

Assumption 5 (Availability). After the death, a qualified set of Registrars, at least one Custodian with a readable copy, and a reading device are available.

Adversaries

We consider the following adversaries, alone or in coalition.

AdversaryDescription
Aout\mathcal{A}_{\mathrm{out}}An outsider who observes the blockchain and submits transactions.
Areg\mathcal{A}_{\mathrm{reg}}A coalition of Registrars, possibly all of them, possibly bribed [3, 27].
Awit\mathcal{A}_{\mathrm{wit}}A wealthy Witness, able to reach the Threshold alone and to create any number of accounts.
Aval\mathcal{A}_{\mathrm{val}}Rational validators who accept bribes to delay or reorder transactions [45, 43, 12].
Acust\mathcal{A}_{\mathrm{cust}}A coalition of Custodians. They have unlimited time with their copy and may digitise it entirely.
Athief\mathcal{A}_{\mathrm{thief}}A party with temporary physical access to a copy, for a time that may exceed τ\tau.
Aenv\mathcal{A}_{\mathrm{env}}An envelope keeper who opens the envelopes.
Adev\mathcal{A}_{\mathrm{dev}}A malicious or compromised reading device.
Aws\mathcal{A}_{\mathrm{ws}}A workshop that digitises the master before destroying it.
Adonor\mathcal{A}_{\mathrm{donor}}A Donor who colludes with an accomplice to defraud Witnesses.
AQ\mathcal{A}_{Q}A future adversary with a quantum computer [19].

Table 4. Adversaries.

Security goals

G1 – Anonymity. Before the Acknowledgment, the link between the civil identity of the Donor and the Donor’s account is hidden.

G2 – No premature inheritance. No inheritance transaction is executed while the Donor is alive.

G3 – Heir exclusivity. The share attached to an heir is paid only to the party to whom the Donor’s arrangements give the corresponding credentials.

G4 – Liveness. After the death, the inheritance is eventually executed.

G5 – Witness fairness. Witnesses who testify truthfully do not lose their Ante.

G6 – Longevity. The above goals are not invalidated by foreseeable cryptanalytic progress over several decades.

TFCP-B

Principle

The Wills are encrypted under a key derived from two halves:

KW=KDF⁡(S∥RW∥id),K_W = \operatorname{KDF}(S \mathbin\Vert R_W \mathbin\Vert id),

where SS is a 256-bit secret shared among the Registrars [40], RW=R(cW)R_W = R(c_W) is the response of the succession key to a public challenge cWc_W, and idid identifies the protocol instance. The Registrars release SS only after the Threshold has been reached and the Deliberation Time has elapsed: SS acts as a time gate. Computing RWR_W requires a copy of the key: it acts as a physical confirmation. The rights of the heirs are attached to other challenges of the same key, which are kept secret.

Security then rests on three independent locks (Table 5), and the third parties involved are powerless: the notary keeps envelopes, the bank keeps an object, the Registrars keep a secret, and none of them holds a key or anything from which a key can be computed.

LockCarried byKept byUseless alone because
Physicala copy of the keyCustodianswithout the challenge, 22562^{256} challenges would have to be tried
Digitala 256-bit challenge in a sealed envelopeenvelope keeper or heirwithout a copy, it designates nothing
Temporalthe AcknowledgmentWitnesses, Registrars, the networkbefore it, an heir’s secret confers no right

Table 5. The three locks.

Setup

  1. Key. A succession key is fabricated in n=3n = 3 to 5 copies and the master is destroyed. The Donor keeps one copy; each of the others is sealed and handed to a Custodian.

  1. Challenges. The Donor draws a public challenge cWc_W, secret challenges c1,…,chc_1,\ldots,c_h uniformly in {0,1}256\{0,1\}^{256} (one per heir slot), a rehearsal challenge c0c_0, and public audit challenges a1,…,a50a_1,\ldots,a_{50}. For each of them the Donor runs Gen on a reading of his own copy and obtains a response and helper data. The workshop takes no part in this step.

  1. Registrars. They are recruited as in TFCP. The Donor draws SS, shares it with a low threshold, and publishes H(S)H(S) and a commitment to each share.

  1. Wills. The Donor publishes, under the identity of the Security Deposit, the ciphertext AEK⁡W(link,signature)\operatorname{AEK}_W(\text{link},\text{signature}) together with idid, cWc_W, the helper data, CW=H(KW)C_W = H(K_W), and the hash of all helper data of the instance.

  1. Slots. For each heir slot ii, the pair (Ci,αi)(C_i,\alpha_i) is recorded, where Ci=H(R(ci))C_i = H(R(c_i)) and αi\alpha_i is the share of the estate. The audit commitments H(R(ay))H(R(a_y)) are recorded as well.

  1. Envelopes. Each cic_i is sealed in an envelope labelled with the name of an heir or with a role, and entrusted to an envelope keeper or given to the heir.

  1. Announcement. It is published as in TFCP.

Acknowledgment

Flowchart showing the TFCP-B process after the death of the Donor

Figure 1. TFCP-B after the death of the Donor.

  1. Witnesses transfer their Antes until the Threshold is reached. Antes are locked until the instance terminates.

  1. During the Deliberation Time TdT_d, a transaction signed by the Security Deposit proves that the Donor is alive. The instance is then aborted and the Antes remain on the Security Deposit.

  1. Otherwise the Registrars publish their shares; the network checks them against the commitments and pays the Registrars.

  1. Physical confirmation. A Custodian reads RWR_W on a copy, computes KWK_W and publishes it. The network checks that H(KW)=CW\mathrm{H}(K_W) = C_W and enacts the Acknowledgment. If no confirmation is published within a period Tc≫TdT_c \gg T_d, the instance is aborted and the Antes are returned, minus a fee.

  1. The Wills can now be decrypted by anybody, and the heir slots become claimable.

The proof of life differs from that of [32], which requires a move on the Donor’s account. That account is hidden until the Acknowledgment, and [32] does not say how the network observes it. If the Registrars decrypt the Wills in order to learn which account to watch, a false alarm discloses the link. A signature by the Security Deposit, which is public, avoids this. The consequences of this choice are analysed in § 7.6.

Claiming a slot

After the Acknowledgment, an heir obtains access to a copy and opens his envelope. He reads Ri=R(ci)R_i = R(c_i) and claims in two transactions:

  1. commit: publish T=H(Ri∥addr∥nonce∥id)T = \mathrm{H}(R_i \mathbin\Vert addr \mathbin\Vert nonce \mathbin\Vert id), where addraddr is an address created for the occasion;

  1. reveal: after Δ\Delta blocks, publish (Ri,addr,nonce)(R_i, addr, nonce). The contract checks that H(Ri)=Ci\mathrm{H}(R_i) = C_i and that a matching commitment is at least Δ\Delta blocks old, and pays αi\alpha_i to addraddr.

A slot that is not claimed within a period TrT_r follows a default rule set by the Donor. This is a hash-locked payment in the sense of [31], preceded by a commitment that prevents a party who observes the reveal from claiming first.

The heir needs no prior key, wallet or presence on the blockchain, and the claim involves no signature scheme other than the one protecting his newly created address. A single key serves all heirs, since by B4 reading one slot gives no information on the others. On a blockchain without such contracts, the slot is replaced by a public key derived from RiR_i and by a time-locked recovery path in the manner of [46].

Rehearsal and proof of custody

The rehearsal slot, of small value, can be claimed while the Donor is alive. It lets the heirs test the copy, the reading device, the envelope and the procedure without revealing anything about the real slots. Every year yy, each Custodian reads R(ay)R(a_y) and publishes it; the network checks it against the recorded commitment. A failing Custodian is thus detected while the Donor can still replace him. These two mechanisms address the absence of feedback noted in Section 4.6 of [33].

From TFCP to TFCP-B: the changes and their rationale

This section makes explicit what TFCP-B changes with respect to TFCP, why, and at what price. Attack numbers refer to Table 7 and Table 8 of the next section.

Attack modelGoalTFCP [32]TFCP-B
A1Collusion of RegistrarsG1Broken. Registrars can pool their Shares at any time; acknowledged in [32].Resists (Claim 1).
A2Bribery of Registrars [3, 27]G1Broken as soon as the bribe exceeds the expected loss of Bail; the collusion is undetectable.Resists unless a Custodian is bribed as well.
A3Paid false alarm to disclose the linkG1Broken if the Registrars decrypt in order to observe the Donor’s account; cost: one Threshold.Resists: SS alone discloses nothing.
A4Chain analysis of the Security Deposit [28]G1Outside the protocol: the Security Deposit must be funded without an on-chain link to the Donor.Same.
A5Wealthy Witness, Donor incapacitatedG2Broken; noted in [32].Resists unless the adversary also has access to a copy (Claim 3).
A6Flash-loan funding of the Threshold [35]G2Not applicable if Antes stay locked beyond one transaction.Same.
A7Sybil WitnessesG2Irrelevant for a stake-weighted Threshold; defeats the countermeasure of [32] based on several separate transactions.The physical confirmation cannot be multiplied.
A8Censorship of the proof of life [45, 43]G2Vulnerable: wrongful Acknowledgment if the Donor’s transaction is censored during TdT_d.Partial: the adversary also needs a copy; a malicious Custodian qualifies.
A9Shallow account, colluding DonorG5Vulnerable; noted in [32].Vulnerable.
A10Theft of the proof-of-life keyG4, G5The key is the Donor’s key: the thief can take the assets directly.Vulnerable: the thief can veto after the death and capture the Antes (§ 7.6).
A11Registrars withhold or corrupt SharesG4Deterred by the Bail; requires a qualified set over decades.Same deterrent; commitments make corruption detectable; a low threshold is affordable.

Table 7. Attacks on the acknowledgment process, and resistance of TFCP and TFCP-B.

Attack modelGoalTFCP [32]TFCP-B
A12 Loss or destruction of all copiesG4Not applicable.Vulnerable; mitigated by nn sites, yearly proof of custody and a time-locked fallback.
A13 Digitisation of a copy by a thief or a CustodianG1, G3Not applicable.Resists (Claims 2, 4).
A14 Envelope keeper colluding with a holder of a copyG3Not applicable.Vulnerable after the Acknowledgment; nothing before (Claim 5).
A15 Malicious reading deviceG3Not applicable.Vulnerable at claim time (§ 7.7).
A16 Front-running of a claim [12, 16]G3Not specified.Resists up to censorship for Δ\Delta blocks.
A17 Helper-data manipulation [13]G3, G4Not applicable.Resists: helper data are bound by an on-chain hash.
A18 Modelling attack on the key [37]G1, G3Not applicable.Resists if B4 holds; to be validated.
A19 Malicious workshopG1, G3Not applicable.Resists: no more than a Custodian.
A20 Quantum adversary [19]G6Signatures and Share distribution are exposed.Slots and Wills rely on hashing with 256-bit secrets; the proof-of-life signature must be migrated by the living Donor.

Table 8. Attacks on custody, claims and the physical key, and long-term attacks.

What is kept

The economic core of TFCP is unchanged: the Donor and the Security Deposit, the Witnesses and their Antes, the Threshold, the Deliberation Time, the Registrars with their Bail and their fees, the encrypted Wills and the public Announcement. TFCP-B remains independent of the underlying blockchain, and an instance that never reaches its Threshold costs nothing more than in TFCP.

The two protocols step by step

StepTFCP [32]TFCP-B
Setup: secretsShares of the key of the Wills are given to the Registrarsshares of SS are given to the Registrars, with public commitments; copies of the succession key are given to the Custodians
Setup: Willsencrypted under the key shared among the Registrarsencrypted under KDF⁡(S∣RW∣id)\operatorname{KDF}(S \mathbin{|} R_W \mathbin{|} id); H(KW)H(K_W) is published
Setup: heirsnot specifiedslots (Ci,αi)(C_i,\alpha_i) are recorded; challenges are sealed in envelopes
Triggerthe Antes reach the Thresholdunchanged
Proof of lifea move on the Donor’s account, which is hiddena signature by the Security Deposit, which is public
End of the Deliberation Timethe Registrars decrypt the Wills; the Acknowledgment followsthe Registrars publish SS; a Custodian publishes KWK_W; the Acknowledgment follows; abort after TcT_c otherwise
Inheritancetransactions are triggered; their content is out of scopeheirs claim their slots by commit and reveal
During the Donor’s lifeno feedbackrehearsal slot; yearly proof of custody

Table 6. TFCP and TFCP-B, step by step.

Seven changes

C1 – Two-factor decryption of the Wills. Why. In TFCP the Registrars can pool their Shares at any time and learn which account belongs to the Donor (attacks A1 and A2). What. The key of the Wills is derived from the Registrars’ secret SS and from a response RWR_W of the succession key. The Registrars hold a time gate, not a decryption key. Price. A copy of the succession key must be readable after the death (A12).

C2 – Proof of life given by the Security Deposit. Why. TFCP requires a move on an account that the network cannot identify before the Acknowledgment; if the Registrars decrypt the Wills to identify it, a false alarm discloses the link (A3). What. A transaction signed by the Security Deposit, which is public, proves life. Price. The proof-of-life key is no longer the key that controls the assets. Its theft allows a third party to abort instances after the death and to collect the Antes (A10).

C3 – Physical confirmation before the Acknowledgment. Why. In TFCP, silence during the Deliberation Time is sufficient. A Donor who is in hospital, or whose transaction is censored, is declared dead (A5, A8). What. After the Registrars have published SS, a holder of a copy must publish KWK_W. Without it the instance is aborted after TcT_c and the Antes are returned, minus a fee. Effect. The Donor is no longer obliged to watch the blockchain permanently: an unanswered false alarm fails by default. Price. The Acknowledgment takes longer and depends on a Custodian. A Custodian who is also an heir gains no protection from this step against himself (A8).

C4 – Heir slots. Why. TFCP stops at the Acknowledgment. The delivery of rights to heirs, in particular to heirs who are under age, have no wallet or are not known in advance, was an open question of [32, 33]. What. Each heir is represented by a hash commitment; the corresponding secret is computed from a copy of the key and a challenge kept in an envelope, and is claimed by commit and reveal. Price. The separation assumption (Assumption 4) and the trust placed in the reading device at claim time (A14, A15).

C5 – Public commitments. Why. A Registrar could publish a wrong Share, and public helper data could be modified (A11, A17). What. Commitments to SS, to each share and to the helper data are recorded at setup, so that any deviation is detected. Price. A few additional hashes on-chain.

C6 – Feedback during the Donor’s life. Why. In TFCP, as in any inheritance scheme, the customer is not the end user, and nothing can be tested before the death [33]. What. A rehearsal slot lets the heirs practise the claim; a yearly proof of custody shows that each Custodian still holds a readable copy. Price. One reading per Custodian and per year, and a negligible part of the capacity of the key.

C7 – Hash-based credentials. Why. An inheritance may be executed decades after its setup, and the signature schemes of today may not survive (A20). What. Slots, confirmation and commitments rely on hashing with 256-bit secrets. The heirs hold no key that could become obsolete. Price. Slots require a blockchain that can check hash preimages and timing conditions; otherwise the time-lock variant of § 8 is used.

Net effect

In TFCP, anonymity and correct execution rest on one population, the Registrars, which is recruited on an open market and assumed to be honest but curious for decades. TFCP-B replaces this single assumption by a conjunction: the Registrars’ secret, a copy of the succession key, and, for the funds, an envelope. No party of a single kind can break anonymity or divert a share. The counterpart is a set of organisational assumptions that TFCP did not need, on Custodians, envelope keepers and reading devices, and the physical logistics that come with them.

Security analysis

We first state what TFCP-B guarantees under the assumptions of § 4, then review attack models from the literature. The arguments are informal.

Guarantees

Claim 1 (Anonymity against Registrars). Under Assumptions 1 and 3, a coalition of all Registrars and outsiders that has no access to a copy learns nothing about the link before the Acknowledgment.

Argument. The coalition knows SS, cWc_W and the helper data. By B2, RWR_W retains close to 256 bits of entropy given the helper data, so KWK_W is indistinguishable from random and the ciphertext hides the link.

Claim 2 (Anonymity against Custodians). A coalition of all Custodians, of AthiefA_{\mathrm{thief}} and of AwsA_{\mathrm{ws}}, holding fewer shares than the threshold, learns nothing about the link before the Acknowledgment, even if every copy is fully digitised.

Argument. The coalition can compute RWR_W but has no information on SS.

Claim 3 (Physical confirmation). No Acknowledgment is enacted unless a party with access to a copy, or to a digitised copy, takes part after the release of SS.

Argument. Enacting requires a preimage of CWC_W, hence both SS and RWR_W.

Claim 4 (Heir exclusivity). Under Assumptions 1–4, and provided the reveal transaction of the legitimate claimant is not censored for Δ\Delta blocks, slot ii is paid to the holder of envelope ii who has access to a copy.

Argument. A valid claim requires RiR_i. Without cic_i, even a fully digitised copy leaves 22562^{256} candidate challenges to test against CiC_i. Without a copy, cic_i gives nothing. A party who learns RiR_i from the reveal must first publish its own commitment and wait Δ\Delta blocks.

Claim 5 (No value before the Acknowledgment). Whatever the coalition, knowledge of RiR_i before the Acknowledgment gives no control over the estate.

Argument. Slots are claimable only after the Acknowledgment, which is enforced by the contract and not by secrecy.

Attack models and resistance

Tables 7 and 8 summarise the analysis; the following subsections discuss the cases that require more than one line.

Collusion, bribery and false alarms (A1–A3, A5)

In TFCP the Shares held by the Registrars are sufficient to decrypt the Wills. The anonymity property is therefore stated in [32] under an honest-but-curious assumption. Registrars are recruited on an open market and serve for decades; pooling Shares leaves no trace, and the literature on bribery shows that such collusions can be organised and paid for trustlessly [3, 27]. We consider that G1 does not hold for TFCP against AregA_{\mathrm{reg}}.

In TFCP-B the same coalition obtains SS and nothing else (Claim 1). A further consequence is that the task of a Registrar can be transferred, and SS re-shared, without risk for G1. The availability of Registrars over decades, an open question of [32], can then be addressed by recruiting many of them with a low threshold.

A false alarm, in which AwitA_{\mathrm{wit}} reaches the Threshold although the Donor is alive, has two possible purposes. The first is to force the disclosure of the link (A3); in TFCP-B, even if the Donor fails to react and the Registrars release SS, nothing is disclosed. The second is to obtain a wrongful Acknowledgment while the Donor is unable to react (A5), which [32] identifies as dangerous. In TFCP-B the Custodians, who are close to the Donor, simply do not confirm; the instance is aborted after TcT_c. The Donor is no longer required to watch the blockchain permanently, which was the main drawback of dead man’s switches. The residual case is an adversary who combines the Threshold, an incapacitated Donor and access to a copy.

Flash loans and Sybil identities (A6, A7)

The original paper mentions a “flash loan attacker” who would reach the Threshold alone. A flash loan is valid within a single transaction and must be repaid by the end of that transaction [35]. An Ante that remains locked on the Security Deposit for the Deliberation Time cannot be financed in this way, so this attack does not apply, provided that the implementation never refunds an Ante within the transaction that deposits it. What remains is the wealthy Witness. The countermeasure suggested in [32], namely to require several separate transactions, is ineffective against an adversary who creates several accounts, unless the transactions are also spread over time, as [32] notes. In TFCP-B the second factor is physical and exists in nn copies only.

Censorship during the Deliberation Time (A8)

Both protocols rely on a time-out: the absence of a proof of life during TdT_d is interpreted as death. Time-outs of this kind are known to be exposed to incentive manipulation. Winzer et al. [45] show that validators can be bribed through a smart contract to ignore a transaction until a time-out elapses, and Tsabary et al. [43] show that such attacks on hashed time-locked contracts are cheaper than previously thought. Here the party with an interest in censoring is whoever benefits from a premature Acknowledgment, in the first place a malicious heir.

TFCP has no other defence than the length of TdT_d and the fee that the Donor is ready to pay. TFCP-B adds the physical confirmation, but this is of limited help when the malicious heir is also a Custodian. We therefore classify the resistance as partial, and recommend (i) a Deliberation Time counted in weeks, since the cost of censorship grows with its duration; (ii) Custodians who have no interest in the estate; and (iii) for large estates, a threshold confirmation in which each Custodian jj is given a personal confirmation challenge and RWR_W is replaced by a secret shared among the corresponding responses, so that kk distinct Custodians must take part.

The proof of life and the shallow-account attack (A9, A10)

In the shallow-account attack of [32], the Donor gives an accomplice the means to prove life. After the death, the accomplice aborts the instance and takes the Antes. No design of the proof of life can prevent this: any secret that proves life can be handed over, and a baryonic key is no exception, since an object does not prove who holds it. Both protocols are vulnerable. The only limits are those mentioned in [32]: the reputation of a named Donor, and the pricing of this risk by Witnesses.

The choice of the key that proves life nevertheless matters. In TFCP it is the key of the Donor’s account. Delegating it amounts to giving away the assets, and its theft is no worse than the theft of the assets. In TFCP-B it is the key of the Security Deposit. This protects anonymity (attack A3), but it creates a key whose theft has a specific effect: after the death of the Donor, the thief can abort every instance and collect the Antes, which violates G4 and G5. We see two mitigations. The key of the Security Deposit can be derived from the succession key and a challenge known to the Donor only, so that it is not stored anywhere. And the time-locked fallback path of Section 5.4 bounds the damage, since it does not depend on the Acknowledgment.

Custodians, envelopes and reading devices (A13–A15)

The design assumes that Custodians, who keep a copy for years, may digitise it entirely: this gives RWR_W, which is useless without SS, and no RiR_i (Claims 2 and 4). The weak form of B3 is therefore sufficient for G1 and G3.

Assumption 4 is organisational, and its violation is the main residual risk for G3. An envelope keeper who opens envelope ii and gains access to a copy, or to a digitised copy, can compute RiR_i. By Claim 5 this is harmless before the Acknowledgment; afterwards the coalition can claim the slot before the legitimate heir. This is the counterpart of the trust placed in the executor of a will, restricted to a diversion after the death. It can be reduced by splitting each challenge between two keepers, and by giving adult heirs their envelope directly.

A reading device sees both the challenge and the response. A malicious device used by an heir can therefore leak RiR_i before the heir has committed, which defeats the commit–reveal step. The device used for claims must be trusted at that moment. We recommend an open design without network interface, and the possibility for an heir to use a device of his own.

Attacks on the physical key (A17–A19)

An adversary who modifies the public helper data can cause Rep to fail or, for some constructions, to leak information [13]. In TFCP-B the hash of all helper data is recorded on-chain at setup, so that modified helper data are rejected.

Modelling attacks [37] exploit the structure of a PUF to predict unseen responses. The relief of a micro-fractured surface has no such structure by design, and B4 states that responses at disjoint sites are independent; an adversary who has scanned part of a copy then knows exactly the responses that fall in that part. B4 is an assumption until measured.

The workshop could measure the master before destroying it. In TFCP-B all responses are generated by the Donor, so the workshop obtains at most what a Custodian with a digitised copy obtains. Combining keys from independent workshops would reduce this assumption further.

Longevity (A20)

Grover’s algorithm [19] reduces the cost of an exhaustive search over NN candidates to about N\sqrt{N} evaluations. With 256-bit challenges and 256-bit responses, searching for a challenge that matches a public commitment, on a digitised copy, costs about 21282^{128} quantum evaluations. The slots, the Wills and the confirmation rely on hashing and symmetric encryption only. The signature of the Security Deposit depends on the signature scheme of the blockchain; a migration can be carried out by the Donor while alive. No action is required from the heirs, who hold no key.

Summary

GoalTFCPTFCP-B
G1 Anonymityonly against honest-but-curious Registrarsholds unless Registrars and a holder of a copy collude
G2 No premature inheritancefails if the Donor cannot react or is censoredrequires in addition access to a copy
G3 Heir exclusivitynot addressedholds under Assumption 4 and with a trusted reading device
G4 LivenessRegistrarsRegistrars, one Custodian, one device; fallback path
G5 Witness fairnessfails against a colluding Donorsame; also fails if the Security Deposit key is stolen
G6 Longevitynot addressedhash-based for heirs; signature migration by the Donor

Table 9. Security goals.

Discussion

The open questions of the earlier papers

Secrets sent to the future ([33], Section 4.2). A slot is defined by a commitment, and the corresponding credentials are an object and a number. The recipient need not exist, be of age, or be known when the slot is created.

Flexibility of wills ([33], Section 4.3). The second paper opposed the rigidity of smart contracts to unforeseen events: an heir who dies first, a collective accident, a change in the law. With slots, the contract fixes shares and names nobody. If the intended heir is no longer there, the envelope follows the ordinary rules of succession and whoever receives it claims the slot. Flexibility is thus returned to the legal system without giving it custody of the funds. Amounts remain fixed, and may conflict with forced-heirship rules [6].

Civil status ([33], Section 4.4). The heir needs no on-chain identity. The link between a person and an envelope is established by the envelope keeper, as it would be for any bequeathed object.

Long term ([33], Section 4.5). The key is passive and its responses are seeds that depend on no particular algorithm. The reading device is the perishable element; its specification has to be open and archived.

Feedback ([33], Section 4.6). The rehearsal slot and the yearly proof of custody provide feedback during the Donor’s life.

Atomicity ([33], Section 4.7). A single key can carry slots for several blockchains. Moreover, CWC_W can be recorded on several blockchains, the publication of KWK_W on one of them serving as a death signal on the others, as the preimage does in an atomic swap. On a blockchain where no Ante is at stake, Registrars colluding with a Custodian could raise this signal cheaply; a local Deliberation Time remains necessary there.

Registrars ([32], Section 5.1). Their task can be transferred and their number increased without consequence for anonymity. Whether a market for Registrars is needed is unchanged.

Minors and unknown heirs ([32], Section 5.1). The copy waits in a vault and the envelope at the notary’s. Neither the guardian nor the bank can spend alone. Envelopes can be labelled with roles, and a residual slot can cover unforeseen cases.

Variants

VariantPrincipleUse
Time lock onlyno distributed Acknowledgment; the heir’s key, derived from the copy, becomes valid after a period of inactivityblockchains without the required contracts
No RegistrarsKW=KDF⁡(RW∥id)K_W = \operatorname{KDF}(R_W \parallel id); Custodians can read the Willssmall estates
Registrars with keysSS is derived from baryonic keys held by the Registrars: no digital Share at restfew institutional Registrars
Threshold confirmationkk distinct Custodians must confirmlarge estates (attack A8)

Table 10. Variants of TFCP-B.

Costs

Table 11 gives the prices of existing services and components that we could verify at the time of writing.

ItemPriceSource
Casa Standard (3-key vault, inheritance)USD 250 per year[5]
Casa Premium (5-key vault)USD 2,100 per year[5]
Casa Private ClientUSD 7,500 per year[5]
Nunchuk, plan with inheritanceUSD 480 per yearas reported in [42]
Liana (software)free, open source[46]
Optical pickup unit (CD/DVD/Blu-ray)about USD 10 per piece[38]

Table 11. Verified market prices (November 2026).

For the baryonic key itself there is no market price, and the following figures are our estimates. The optical head is a commodity (Table 11), and polished borosilicate substrates of 100 mm are sold by laboratory suppliers for a few tens of dollars per unit; the cost is dominated by the replication, coating and sealing steps, performed in a small trusted facility. We estimate a set of three copies at EUR 95–140 and a set of five at EUR 140–220 in small series, a prototype reading device at EUR 2,000–5,000, and a device produced in series at EUR 150–400. A device kept by a notary or a bank would serve all their clients.

256-bit response, 8 % error margin122 sites, read in about 35 s
Kit for 11 heirs63 responses: 1 for RWR_W, 11 slots, 1 rehearsal, 50 audits
Surface used7,686 sites, i.e. about 0.01 % of one side
Readings at setupabout 37 minutes
Claim by an heirone reading, two transactions

Table 12. Sizing of a succession kit under the estimates of Table 1.

Over thirty years, the least expensive assisted service of Table 11 amounts to USD 7,500, and it depends on the continued existence of its provider. A baryonic key is a one-off expense, but it requires access to a reading device when the time comes.

Limitations

  • No experimental validation. The candidate instantiation has not been built. The raw error rate between copies, the entropy per site, the independence of sites, ageing, and the time needed to digitise a copy are estimates. If τ\tau turned out to be much shorter, the analysis of § 7 would still hold, since it assumes that copies may be digitised; if the error rate were much higher, capacity and reading time would degrade.

  • Informal analysis. We have given arguments, not proofs. A formal model of the protocol, including the game-theoretic behaviour of Witnesses, Registrars and validators, remains to be written. The parameters TdT_d, TcT_c, TrT_r, Δ\Delta, the Threshold, the Antes and the Bails have to be set, and as noted in [32] only practice will settle some of them.

  • Attacks that remain. The shallow-account attack (A9), censorship during the Deliberation Time (A8), the theft of the proof-of-life key (A10), the violation of the separation assumption (A14) and a malicious reading device (A15) are not solved.

  • The oracle does not prove death. The Acknowledgment remains an economic bet completed by a physical gesture. It also makes a payment conditional on a death, so the fees paid to Witnesses must remain small enough not to create an incentive to cause it.

  • Availability. Enough Registrars, one readable copy and one device are needed after the death.

  • Re-centralisation. Vaults and notaries are back in the loop. They have no power over the funds, but they can fail in their custody.

  • Law. Jurisdiction, taxation and forced heirship are not addressed.

Conclusion

The second paper of this series suggested that the materiality of an object could work around the perfect and non-destructive copy that digital technologies allow. Baryonic keys make this suggestion precise: keys cloned a small, fixed number of times, whose bits are never stored. The instantiation that we outlined makes digitisation slow and visible rather than impossible, and the protocol was designed accordingly, so that a digitised copy gives nothing in isolation.

Applied to the Tales From the Crypt Protocol, the primitive removes the dependence of anonymity on the honesty of Registrars, blocks the inheritance of an incapacitated Donor after a false alarm, and provides what the protocol had left out of scope: a way to deliver rights to heirs who hold no key, are under age, or are not yet known. The analysis also shows what it does not provide. The shallow-account attack is untouched, censorship during the Deliberation Time remains a concern, and custodians, envelope keepers and reading devices bring assumptions of their own.

Future work is experimental, on the physical key, and formal, on the protocol. We believe that keys which exist in a few places and nowhere else are of interest beyond inheritance; the uses outlined in § 3.4 deserve an analysis of their own.

References

  1. [1][BCD+23] Alysson Bessani, Miguel Correia, Tobias Distler, Rüdiger Kapitza, Paulo Esteves-Verissimo, and Jiangshan Yu. Vivisecting the dissection: On the role of trusted components in BFT protocols. arXiv:2312.05714, 2023.arxiv.org/abs/2312.05714
  2. [2]Keith Bonawitz, Vladimir Ivanov, Ben Kreuter, Antonio Marcedone, H. Brendan McMahan, Sarvar Patel, Daniel Ramage, Aaron Segal, and Karn Seth. Practical secure aggregation for privacy-preserving machine learning. In ACM Conference on Computer and Communications Security (CCS), 2017.DOI
  3. [3]Joseph Bonneau. Why buy when you can rent? Bribery attacks on Bitcoin-style consensus. In Financial Cryptography and Data Security – FC 2016 International Workshops. Springer, 2016.
  4. [4]Xavier Boyen. Reusable cryptographic fuzzy extractors. In 11th ACM Conference on Computer and Communications Security (CCS), pages 82–91. ACM, 2004.
  5. [5]Casa. Pricing. https://casa.io/pricing, accessed November 2026.
  6. [6]Cristina Carata and Ana-Luisa Chelaru. The evolution of the digital inheritance: Legal, technical, and practical dimensions of cryptocurrency transfer through succession in French-inspired legal systems. arXiv:2410.22907, 2024.arxiv.org/abs/2410.22907
  7. [7]Ronald Cramer, Ivan Damgård, and Yuval Ishai. Share conversion, pseudorandom secret-sharing and applications to secure computation. In Theory of Cryptography Conference (TCC), volume 3378 of LNCS, pages 342–362. Springer, 2005.DOI
  8. [8]Stephen Y. Chou, Peter R. Krauss, and Preston J. Renstrom. Imprint lithography with 25-nanometer resolution. Science, 272(5258):85–87, 1996.DOI
  9. [9]Christian Cachin, Klaus Kursawe, and Victor Shoup. Random oracles in Constantinople: Practical asynchronous Byzantine agreement using cryptography. Journal of Cryptology, 2005. Preliminary version in PODC 2000, pages 123–132.
  10. [10]Kevin Choi, Aathira Manoj, and Joseph Bonneau. SoK: Distributed randomness beacons. In IEEE Symposium on Security and Privacy (S&P), 2023.DOI
  11. [11]Sourav Das, Sisi Duan, Shengqi Liu, Atsuki Momose, Ling Ren, and Victor Shoup. Asynchronous consensus without trusted setup or public-key cryptography. In ACM Conference on Computer and Communications Security (CCS), pages 3242–3256, 2024.DOI
  12. [12]Philip Daian, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach, and Ari Juels. Flash boys 2.0: Frontrunning in decentralized exchanges, miner extractable value, and consensus instability. In IEEE Symposium on Security and Privacy (S&P), 2020.DOI
  13. [13]Yevgeniy Dodis, Jonathan Katz, Leonid Reyzin, and Adam Smith. Robust fuzzy extractors and authenticated key agreement from close secrets. In Advances in Cryptology – CRYPTO 2006, volume 4117 of LNCS, pages 232–250. Springer, 2006.
  14. [14]drand: a distributed randomness beacon. https://drand.love/, accessed November 2026.
  15. [15]Yevgeniy Dodis, Leonid Reyzin, and Adam Smith. Fuzzy extractors: How to generate strong keys from biometrics and other noisy data. In Advances in Cryptology – EUROCRYPT 2004, volume 3027 of LNCS, pages 523–540. Springer, 2004. Full version with Rafail Ostrovsky: SIAM Journal on Computing, 38(1):97–139, 2008.
  16. [16]Shayan Eskandari, Seyedehmahsa Moosavi, and Jeremy Clark. SoK: Transparent dishonesty: front-running attacks on blockchain. In Financial Cryptography and Data Security – FC 2019 International Workshops, pages 170–189. Springer, 2019.arxiv.org/abs/1902.05164
  17. [17]Scott Fluhrer, Panos Kampanakis, David McGrew, and Valery Smyslov. Mixing preshared keys in the Internet Key Exchange Protocol Version 2 (IKEv2) for post-quantum security. RFC 8784, IETF, June 2020.
  18. [18]Hanwen Feng and Qiang Tang. Asymptotically optimal adaptive asynchronous common coin and DKG with silent setup. In Advances in Cryptology – CRYPTO 2025. Cryptology ePrint Archive, Paper 2024/2098.DOI
  19. [19]Lov K. Grover. A fast quantum mechanical algorithm for database search. In 28th Annual ACM Symposium on Theory of Computing (STOC), pages 212–219. ACM, 1996.arxiv.org/abs/quant-ph/9605043
  20. [20]Suyash Gupta, Sajjad Rahnama, Shubham Pandey, Natacha Crooks, and Mohammad Sadoghi. Dissecting BFT consensus: In trusted components we trust! In 18th European Conference on Computer Systems (EuroSys). ACM, 2023.arxiv.org/abs/2202.01354
  21. [21]Edwin En-Te Hwu and Anja Boisen. Hacking CD/DVD/Blu-ray for biosensing. ACS Sensors, 2018.DOI
  22. [22]Chelsea Komlo and Ian Goldberg. Arctic: Lightweight and stateless threshold Schnorr signatures. In Public-Key Cryptography – PKC 2025. Springer, 2025.DOI
  23. [23]Eleftherios Kokoris-Kogias, Dahlia Malkhi, and Alexander Spiegelman. Asynchronous distributed key generation for computationally-secure randomness, consensus, and threshold signatures. In ACM Conference on Computer and Communications Security (CCS), pages 1751–1767, 2020.DOI
  24. [24]Harish Karthikeyan and Antigoni Polychroniadou. OPA: One-shot private aggregation with single client interaction and its applications to federated learning. arXiv:2410.22303, 2024.DOI
  25. [25]Anne M. Luescher, Andreas L. Gimpel, Wendelin J. Stark, Reinhard Heckel, and Robert N. Grass. Chemical unclonable functions based on operable random DNA pools. Nature Communications, 15:2955, 2024.DOI
  26. [26]T. McGrath, I. E. Bagci, Z. M. Wang, U. Roedig, and R. J. Young. A PUF taxonomy. Applied Physics Reviews, 6:011303, 2019.DOI
  27. [27]Patrick McCorry, Alexander Hicks, and Sarah Meiklejohn. Smart contracts for bribing miners. In Financial Cryptography and Data Security – FC 2018 International Workshops. Springer, 2018.DOI
  28. [28]Sarah Meiklejohn, Marjori Pomarole, Grant Jordan, Kirill Levchenko, Damon McCoy, Geoffrey M. Voelker, and Stefan Savage. A fistful of bitcoins: Characterizing payments among men with no names. In Internet Measurement Conference (IMC), pages 127–140. ACM, 2013.
  29. [29]Yiping Ma, Jess Woods, Sebastian Angel, Antigoni Polychroniadou, and Tal Rabin. Flamingo: Multi-round single-server secure aggregation with applications to private federated learning. In IEEE Symposium on Security and Privacy (S&P), 2023.DOI
  30. [30]Abraham Pais. On the baryon-meson-photon system. Progress of Theoretical Physics, 10(4):457–469, 1953.DOI
  31. [31]Joseph Poon and Thaddeus Dryja. The Bitcoin Lightning Network: Scalable off-chain instant payments, 2016.
  32. [32]Frédéric Prost. On the heritage of crypto assets – Tales From the Crypt Protocol. arXiv:2209.11194, 2022.
  33. [33]Frédéric Prost. Inheritance and blockchain: Thoughts and open questions. arXiv:2212.01194, 2022.arxiv.org/abs/2212.01194
  34. [34]Ravikanth Pappu, Ben Recht, Jason Taylor, and Neil Gershenfeld. Physical one-way functions. Science, 297(5589):2026–2030, 2002.
  35. [35]Kaihua Qin, Liyi Zhou, Benjamin Livshits, and Arthur Gervais. Attacking the DeFi ecosystem with flash loans for fun and profit. In Financial Cryptography and Data Security – FC 2021, pages 3–32. Springer, 2021.
  36. [36]Michael O. Rabin. Randomized Byzantine generals. In 24th Annual Symposium on Foundations of Computer Science (FOCS), pages 403–409. IEEE, 1983.DOI
  37. [37]Ulrich Rührmair, Frank Sehnke, Jan Sölter, Gideon Dror, Srinivas Devadas, and Jürgen Schmidhub­er. Modeling attacks on physical unclonable functions. In 17th ACM Conference on Computer and Communications Security (CCS), pages 237–249. ACM, 2010.
  38. [38]Rekha S. Sekar and Arvind Ajoy. Synchronous detection to reduce offsets in focus error of an optical pickup unit. IEEE Transactions on Instrumentation and Measurement, 2021. doi:10.1109/TIM.2020.3016072.
  39. [39]Sarcophagus. Sarcophagus – a dead man’s switch. https://medium.com/sarcophagus-community/sarcophagus-a-dead-mans-switch-92e70de66554, accessed November 2026.
  40. [40]Adi Shamir. How to share a secret. Communications of the ACM, 22(11):612–613, 1979.
  41. [41]Valery Smyslov. Mixing preshared keys in the IKE_INTERMEDIATE and CRE-ATE_CHILD_SA exchanges of the Internet Key Exchange Protocol Version 2 (IKEv2) for post-quantum security. RFC 9867, IETF, November 2025.
  42. [42]Spark. Bitcoin inheritance planning tools compared. https://www.spark.money/tools/bitcoin-inheritance-planning-tool-comparison, accessed November 2026.
  43. [43]Itay Tsabary, Matan Yechieli, Alex Manuskin, and Ittay Eyal. MAD-HTLC: Because HTLC is crazy-cheap to attack. In IEEE Symposium on Security and Privacy (S&P), pages 1230–1248, 2021.
  44. [44]Giuliana Santos Veronese, Miguel Correia, Alysson Neves Bessani, Lau Cheuk Lung, and Paulo Verissimo. Efficient Byzantine fault-tolerance. IEEE Transactions on Computers, 62(1):16–30, 2013.
  45. [45]F. Winzer, B. Herd, and S. Faust. Temporary censorship attacks in the presence of rational miners. In IEEE European Symposium on Security and Privacy Workshops (EuroS&PW), pages 357–366, 2019.
  46. [46]Wizardsardine. Liana: a Bitcoin wallet with a time-locked recovery path. https://github.com/wizardsardine/liana, accessed November 2026.

Paper details

vlnc:W_z3b9s3ex
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents