HIPAA-Adjacent Secure Messaging for Healthcare

HIPAA's Security Rule (45 CFR §164.312) requires covered entities to safeguard electronic Protected Health Information (ePHI) in transit — access controls, transmission security, audit controls. Standard encrypted messaging protects message bodies but exposes thecommunication graph: which provider messaged which patient, when a lab result was delivered, which specialist a primary referred to. In healthcare that graph is itself ePHI. It can reveal diagnoses (a patient's oncologist contacted frequently), provider relationships (referral patterns), and patient conditions (who is receiving lab results and for what). Tessera protects both the message and the metadata.

The problem: communication metadata is ePHI

The HHS guidance on "de-identification" (§164.514) acknowledges that derived data can identify an individual — and communication patterns are notoriously re-identifiable. A messaging server that logs provider↔patient and provider↔provider traffic holds a graph from which diagnoses, treatment timelines, and care relationships are inferrable. Encrypting the body does not remove that exposure; it merely narrows it to the routing layer, which is itself a regulated system component under §164.308.

Three common flows each carry metadata risk:

How Tessera helps

Tessera's three privacy mechanisms map onto the properties a covered entity needs for compliant messaging:

• ZK authentication → access controls. A Schnorr / Fiat–Shamir proof under a blinded pseudonym authenticates the sender cryptographically. No shared password, no replayable token; the recipient's verifier accepts or rejects in ~13 ms.
• AES-GCM payload encryption → transmission security. The proof and payload travel inside an AES-GCM ciphertext; the transport never sees plaintext.
• Metadata privacy → narrower ePHI surface. Blinded pseudonyms and bucketed broadcast mean the relay does not hold a sender↔recipient graph. What the relay cannot store, it cannot breach.
• (ε,δ)-DP cover traffic → quantifiable privacy. The cover-traffic budget is a tunable, auditable parameter a security risk analysis (§164.308(a)(1)) can cite.

Architecture

Recommended deployment: the covered entity (or its business associate, under a BAA) self-hosts relay nodes inside its trust boundary. Each provider and patient holds a long-term keypair; enrolment is pairwise (provider↔patient, provider↔provider, lab↔provider). Messages route to recipient buckets; the recipient pulls.

Provider (long-term key)          Patient / second provider (verifier key)
────────────────────             ────────────────────────────────
shared_seed from enrolment        shared_seed from enrolment
  │
  1. BlindedSender.pseudonym()    → Y' (fresh, unlinkable)
  2. ZKProver.prove(message)       → π (Schnorr FS)
  3. SecureEncryption.encrypt()   → AES-GCM ciphertext
  4. broadcast → patient bucket   (relay sees bucket, not recipient)
                                  │
                                  5. pull from bucket
                                  6. decrypt → (π, message)
                                  7. BlindedVerifier.verify(π, Y', message)
                                     → accept: verified provider
                                     → reject: forged / unknown

Compliance mapping (§164.312)

Architectural mapping — not legal advice; a covered entity's security risk analysis must confirm applicability.

Code example

from tessera.sdk import Sender, Verifier
from tessera.crypto.blinding import BlindedSender, BlindedVerifier
from tessera.crypto import SecureEncryption

# enrolment: provider + patient share a seed (in person, or via EHR portal)
shared_seed = os.urandom(32)

# provider sends a follow-up instruction
provider = Sender(secret_key=b"...provider long-term sk...")
bs = BlindedSender(provider.public_key, shared_seed, session_id="visit-7421")
y_prime = bs.pseudonym()
proof   = bs.prove(b"Continue metformin 500mg BID; recheck HbA1c in 3 months.")
ct      = SecureEncryption.encrypt(proof, b"Continue metformin ...", patient_pubkey)

# routed over self-hosted relay inside the health system VPC
relay.broadcast(bucket=compute_bucket(patient_fingerprint), ciphertext=ct)

# patient's device verifies
bv = BlindedVerifier(
    Verifier(expected_pubkey=provider.public_key),
    shared_seed, session_id="visit-7421",
)
assert bv.verify(proof, y_prime, b"Continue metformin 500mg BID ...") is True

Frequently asked questions