K.R.Y.P.T.O.S.
Distinguish valid, malformed, modified and repeated synthetic input envelopes.
Extension of AEGIS / FRONTIER-NET XDP research
What has actually been tested
Candidate against best known reference
Bounded candidate evaluation
Distinguish valid, malformed, modified and repeated synthetic input envelopes.
Optimized incumbent
Optimized incumbent ACL/XDP policy with independent false-drop oracle and equivalent workload.
Interactive receipt verification
This browser only verifies precomputed synthetic evidence. It never runs private algorithms or interacts with real equipment.
Select one of four public reference scenarios.
Loading evidence…
Scenario evidence: four controlled reference cases
Synthetic reference illustrations only. No customer workload, physical actuator or operational network is simulated at real-world fidelity.
Valid envelope
A conforming synthetic envelope satisfies the reference oracle.
Malformed input
Schema mismatch is identified in the toy fixture.
Integrity mismatch
Corruption invalidates the synthetic acceptance.
Repeated sequence
A repeated test sequence is not admitted.
Evidence downloads
All samples are synthetic. No customer's data, private program source, driver code, bitstreams, secret thresholds or deployment details are published.
Industrial acceptance — eight gates
Next evidence needed: Driver-mode support, load tests with packet loss, false-positive rate and real kernel timing.
Diagnostic with a hard boundary
One approved non-production sample, one optimized comparator, up to 30 paired repetitions per supported scenario, written GO / NO-GO / INCONCLUSIVE. Two scheduled calls of 20 minutes each; everything else asynchronous.
Target seven business days from complete approved inputs; 50% at signature and 50% at delivery; taxes in signed quote. NO-GO is a complete result. No PoV, source transfer, support retainer or production modification included.
Request the technical diagnostic
Send only a business email, scale and environment. Never submit proprietary files through this form.
Security and physical evidence — 13 public profiles
OWASP ZAP 2.17.0, seven completed passive spider scans. Recorded: 1 High (SHA-256 digits, false-positive candidate), 89 Medium (public CORS header observations), 217 Informational. Original findings retained; no active pentest certification.
OVHcloud Manager authenticated and FPGA Pi3 laboratory web page reports connected control service and power on. No new authenticated SSH/JTAG test, electrical timing or hardware NIC offload was run from this session. Earlier Linux reference tests are separate and narrowly scoped.
No private source, bitstreams, keys, customer traces or industrial control commands are published.
NIC hardware timestamping - 64/64 paired
Operator-reported live OVH run: 64/64 raw hardware RX and TX timestamps matched on one NIC. p50: 136,837 ns (136.837 µs). p99: 163,625 ns (163.625 µs). Original NIC configuration restored.
Boundary: this is neither sub-microsecond nor network wire-to-wire. It does not benchmark any of the seven reference profiles, private motors or FPGA logic; 64 observations do not establish production tail percentiles.