Resources Open Doors S.A.S.Comparative presentationGrey-box editionLeer en español

La Familia

Four ways for a message not to exist.

Abisal

Two passphrases. The artwork is the message. The balance between strength and friction.

Críptida

Four passphrases. The artwork is the message. The most ambitious and best measured cryptographic design in the house.

Filigrana

Four passphrases. The artwork is the message. The cipher where the data picks its own route.

CipherVault

One passphrase. The message lives inside an image that no reference forensic tool tells apart from an empty one. And it can be published.

Common ground

What the four of them share

It is worth saying first, because it is what makes the differences comparable.

Standard core: In all four, the lock on the content is authenticated AES-GCM encryption run by the browser. None of them replaces that core with an invention of our own; whatever is our own design always sits on top of it.
Memory-hard derivation: Passphrases are never used as keys: they go through Argon2id, which forces hundreds of MiB to be reserved for every guess and ruins the advantage of specialised hardware.
Authentication in the right order: They verify in constant time and fail cleanly: a wrong passphrase produces an explicit error, never corrupt text handed over as valid.
Entirely client-side: No servers, no remote accounts, no telemetry, no downloads at run time.
PNG written by hand: They read and write the file without delegating to the browser, because anti-fingerprinting defences alter pixel readings and would corrupt the data silently.
No recovery: There is no back door, no backup and no third-party custody. Losing the passphrases is losing the content.

The fork

One question splits the family in two

And it almost always decides the recommendation: does a clean version of the image exist to compare it against?

Carrier-free · Abisal, Críptida, Filigrana

The image does not hide the message: it is synthesised from it. Every pixel is payload. There is no original, so there is no differential to analyse: a whole family of steganalysis techniques is structurally out of play.

The price is that the artwork is 100 % synthetic, high-entropy and of fixed size — and that, to an analyst who already suspects the channel, is a trait in itself.

With carrier · CipherVault

The message is embedded in an image that exists: art the system itself generates, or an image the user supplies. In principle, a version without the message exists.

The defence is that the two are indistinguishable — and that indistinguishability, unlike the argument above, has been measured against the reference forensic tools and against deep-learning detectors.

The underlying difference between the two branches is not how much they hide, but what kind of guarantee they offer: the carrier-free branch gives a structural argument, solid and unmeasured. CipherVault gives empirical evidence against the detectors that exist today.

The four

Each one, and the market it goes to

Abisal the balance

What defines it: Two passphrases instead of four, and an outer layer that diffuses at multiple scales in both directions, with a final stage of secret size that depends on the encrypted data itself. It keeps real defence in depth — two independent costly derivations — with half the ceremony.
Recommended use: Sensitive personal communication; custody of your own seed phrases and credentials; notes that have to live on a device without looking like what they are. Two passphrases also work as custody between two people: a couple, two partners.
What it gives up: Lower total brute-force cost than the four-passphrase variants, and no ability to split custody among more than two parties.

Críptida the best design

What defines it: Four passphrases, and an idea with no known equivalent: the message itself decides which passphrase keys each round of the cipher. The assignment is secret, data-dependent and irregular. The structure rests on a Feistel network, for which a classical theorem backs the construction, with a round margin two orders of magnitude above the required minimum.
Recommended use: Any secret that must not be openable by a single person: treasury keys split among four signatories, succession instructions, shareholder agreements. It is the only one in the carrier-free branch that encrypts at full capacity in the order of a second.
What it gives up: The four-passphrase ceremony is heavy for individual use, and 4-of-4 custody is strict unanimity: the loss or bad faith of a single custodian makes the content unrecoverable.

Filigrana the most conservative topology

What defines it: Also four passphrases, but an outer cipher of a different nature: instead of fixed rounds, the data picks its route twice per byte — which passphrase keys it, and which of three qualitatively distinct circuits transforms it. It is not the same operation under another key: jumping between operations changes the whole circuit.
Recommended use: A seed phrase, a recovery key, an instruction a few lines long, where performance does not matter and the double data dependency pays off most. As a desktop piece, the engraving also works as an object in its own right.
What it gives up: Performance, and not marginally: the routing is sequential by construction, and at full capacity encryption is counted in minutes, not seconds. That is the price of having the circuit depend on the data.

CipherVault the product

What defines it: A single passphrase, an entirely standard core, and all the intelligence spent on the hiding: adaptive coding that picks, among all the ways of hiding the same message, the one of lowest detectable cost; a sweep of the low bit plane in two passes, so the footprint does not depend on message length; constant file weight; transparency channel untouched; zero markers.
Recommended use: When the image has to survive analysis, not just a glance; when the message exceeds what fits on a fixed canvas; when the tool is deployed for several people with different permissions; when the artwork is published.
What it gives up: There is no secret sharing: a single passphrase governs the message. An original of the image exists. And if the file passes through a channel that recompresses to a lossy format, the message dies.

Comparison

Side by side

The four systems, axis by axis
	Abisal	Críptida	Filigrana	CipherVault
Approach	Carrier-free	Carrier-free	Carrier-free	With carrier
Secret passphrases	2	4	4	1 (+ optional file)
Custody model	2-of-2	4-of-4	4-of-4	Individual
Cryptographic core	AES-GCM	AES-GCM	AES-GCM	AES-256-GCM
Own outer layer	MAREA	TELAR	Lanzadera	None load-bearing
Critical path	Own, over standard	Own, over standard	Own, over standard	100 % standard
Capacity	~32 / 65 KB, fixed	~32 / 65 KB, fixed	~32 / 65 KB, fixed	Scales with resolution
Speed at full capacity	~9 s	~1 s	~100 s	Fast
Evidence of hiding	Structural argument	Structural argument	Structural argument	Measured
Plausible deniability	Implicit	Implicit	Implicit	Operational
Aesthetic	Abyssal water	Jewelled glass	Engraving / oil	Free
Ecosystem	Single file	Single file	Single file, desktop	NFT, gallery, multi-user
Friction of use	Medium	High	High	Low

Decision tree

If the client…

Cannot rule out that the adversary obtains the originalCríptida

Needs the secret not to be openable by one person aloneCríptida · Abisal

Wants serious protection with minimal ceremony, carrier-freeAbisal

Needs the image to survive a dedicated forensic analysisCipherVault

Has messages that do not fit on a fixed canvasCipherVault

Will deploy the tool for a team with permissionsCipherVault

Requires that no unaudited cipher hold the confidentialityCipherVault

Keeps a few lines of very high value and time does not matterFiligrana

Wants the artwork as an object, on the desktopFiligrana

Nothing dressed up

The limits, which are part of the product too

Filigrana and Críptida compete for the same client. They share passphrases, custody, capacity, core and the whole steganographic machinery. They differ in the outer cipher, and there Críptida is about a hundred times faster and has a considerably more extensive validation analysis. The argument in favour of Filigrana is qualitative and is not quantified today.

Abisal does not compete with Críptida: it tiers it. Two passphrases against four is a deliberate product decision, not a reduced version.

The proprietary layers have not been audited by third parties. The three carrier-free systems stack a layer of our own on top of the standard core, and that layer has not been through review by cryptographers. The stacking never weakens anything: it adds. In CipherVault the discipline is stricter, because every proprietary transformation is applied to data that is already encrypted and none of it carries weight.

Hiding the construction is not a security measure. A system must be secure even if everything except the key is public. The reserve of this page is commercial, not cryptographic.

Kerckhoffs's principle

White editions

Audit is welcome

This page is grey box: it compares the four systems by what they do, by the properties that were measured and by the limits of each one, and it holds back the recipes — the parameters of the outer ciphers, the mixing formulas, the selection rules and the arrangements that determine where each piece of data lives inside the image.

The white editions exist: the complete technical specification of each system, with parameters, constructions and measurements in detail. They are available under agreement for cryptographic review, together with the commercial licence. It is the only way a proprietary design earns trust, and there is no short path.

ingenieriaaplicada@opendoors.com.ar

Resources Open Doors S.A.S. · Mendoza, Argentina