Map crypto to flows
Import CBOMs from any CycloneDX tool. Every link of a UPI, AePS or eKYC flow shows its key exchange, certificate and owner.
Oversized ClientHello
A 1,488-byte hybrid hello spans two packets; old inspection boxes reset it.
Prove before
you switch.
A digital twin of your payment and identity flows. Run real ML-KEM and ML-DSA through it on Indian networks, against your partners and HSMs, before anything reaches production.
01 · Engine
Four engines, one loop.
Find where quantum-vulnerable cryptography sits on each business flow, prove a change on a twin, fix what breaks, and keep evidence of every step. Then do it again when the next algorithm arrives.
Test on a twin
Classical, hybrid, ML-DSA, proxy and pure PQC, run through eight chaos tests and six Bharat network profiles.
Turn failures into diffs
Each failure comes with a plain explanation, configuration diffs and a re-test, then a guarded canary in production.
Hand over proof
Risk in rupees for the board. Records signed with Ed25519 and ML-DSA-65 that auditors verify offline.
02 · Twin
Twin one flow. Leave production alone.
The twin copies a single business flow, from gateway and PSP to switch, issuer, HSMs and the inspection box in between, with your real versions. The rest of the estate is untouched while you test.
98 ms p99 · X25519MLKEM768
03 · Chaos tests
Eight ways a migration breaks.
Each test reproduces a failure teams have already hit in real rollouts. The twin runs all eight against every option, on every hop, before you change a single production setting.
CH-01
CH-02
MTU black holes
Tunnels and cellular uplinks drop full-size packets when ICMP is blocked.
CH-03
Large certificate chains
ML-DSA certificates push the server reply to 15 KB, past an 8 KB POS SDK buffer.
CH-04
HSM ceilings
Firmware without ML-DSA, or signing capacity that runs out at UPI peak.
CH-05
Silent fallback
A load balancer quietly negotiates classical groups; nothing fails, nothing is safe.
CH-06
Rotation storms
New certificates flush sessions and break apps that pin the old hierarchy.
CH-07
Lagging partners
A PSP or switch on Java 17 cannot verify your new ML-DSA certificate.
CH-08
Bharat networks
Fibre to 2G/EDGE and VSAT: loss and latency that multiply larger handshakes.
04 · Evidence
Inventories are crowded. Proof is missing.
Discovery tools tell you what is vulnerable. Regulators and boards now ask whether the fix works, in your own flows, with your partners. That answer needs a test and a record nobody can quietly edit.
7%
of organisations have deployed post-quantum cryptography widely, while 87% are planning it (DigiCert, May 2026, n=1,001).
2.4/5
average quantum readiness rated by 118 financial-sector CISOs and CTOs in India (ISB study).
2029
target for migrating critical information infrastructure, in the DST roadmap of February 2026.
1,488 B
hybrid ClientHello measured with OpenSSL 3.5, against 312 bytes for classical X25519.
05 · Questions
What people ask first.
If your question is about a specific flow, partner or HSM, the demo console answers it with a run you can inspect line by line.
Do you touch production keys or traffic?
No. CryptoTwin reads exported CBOMs, configuration bundles and handshake metadata. It never holds a production private key and never sees payloads. Tests run on the twin; only the guarded rollout touches production, through your own load balancer.
Is the UPI flow in the twin the real one?
It is a simulation built from public descriptions of UPI-style, AePS-style and eKYC-style flows. It does not use NPCI or UIDAI specifications and never connects to their systems. Your own components run with your real versions.
What does “real cryptography” mean here?
ML-KEM and ML-DSA run through OpenSSL in every test, and handshake sizes come from live TLS 1.3 handshakes on OpenSSL 3.5. The network around them is simulated with the Bharat profiles, or applied with tc/netem when the twin runs on Kubernetes.
Can an auditor check the evidence without trusting you?
Yes. Records are hash-chained, signed with Ed25519 and ML-DSA-65, and committed to a Merkle log. An auditor verifies an exported bundle with one Python file and one library, offline.
How can a small co-operative bank afford this?
Through a shared twin. The ASP or sponsor bank that runs their core banking tests the shared stack once; each member bank gets its own variant and evidence pack, for ₹2–5 lakh a year.
What about Thales, Entrust or JISA HSMs?
The HSM lab models each class of device by firmware and mechanism. The demo profiles are placeholders; at a pilot they are replaced by measurements taken from the vendor's dev kit over PKCS#11.