NETRA reads a recorded packet capture and tells you how good the encryption on your email traffic actually was. It never contacts any server. This page walks through every screen, in the order you would use them.
1 · Pick a capture
Library
Every capture available to analyse. The shipped ones demonstrate different
conditions — a legacy server negotiating TLS 1.0, a healthy modern exchange,
a large mixed enterprise capture. Pick one, or upload your own
.pcap.Nothing you upload leaves this machine, and nothing is stored
after the analysis finishes. NETRA holds one capture in memory, reports, and
forgets it.
2 · Run it
Analysis
Start the run and watch each stage report as it completes — reassembly,
protocol identification, STARTTLS, TLS facts, certificates, then the reasoning
layer. This is streaming from the server as it actually executes, not an
animation.
3 · Read the result, top to bottom
Capture Health Passport
Read this first. Before saying anything about security, NETRA reports how good
the evidence was: how many frames decoded, whether packets were truncated,
whether the capture began mid-conversation. A tool that leads with a score and
hides its limitations has chosen which one it wants you to believe.
Observed risk & Evidence coverage
Two numbers, never combined. Observed risk is how bad what we saw
was. Evidence coverage is how much of the picture we got.
Missing evidence lowers coverage and leaves risk untouched — so a capture we
barely saw can never be mistaken for a server that is safe.
Harvest-now-decrypt-later exposure
The share of sessions whose key exchange could be broken later by a quantum
computer. Note that it also shows how many sessions were undeterminable —
those are excluded from both sides of the calculation rather than quietly
counted as safe.
Server support matrix
One card per server, pooling every session that reached it. A capability marked
DEMONSTRATED was selected by the server, which is proof. One
marked EXCLUDED_FIRM was offered and refused, and we know the
server's preference order, so the refusal is conclusive.
UNDETERMINED means nobody in the capture ever asked.
Click any evidence link to jump to the frame that established that cell.
4 · Go into a session
Sessions → click any row
The drawer opens with every fact this conversation produced, each carrying a
badge saying how it is known: read from bytes, proved across sessions, predicted
by a model, or not observable. Where something is unobservable, the reason is
given too — for instance
midstream_capture, meaning the recording
started after the handshake.Negotiation signals
Shows how many version steps below the client's offer the server settled, and
states plainly whether any corroborating downgrade sentinel or fallback signal
was present. A large gap is consistent with an attack and equally consistent
with an old server, and NETRA will not choose between them without evidence.
Packet Ladder
The bottom of the drawer, and the end of the evidence chain. Client on the left,
server on the right, time running down. Findings are attached to the frames that
produced them — hover any frame for its length and layer. You are not told
there is a problem somewhere; you are shown the packet where the server made the
choice.
5 · Decide what to fix
Fix-first
The same findings, reordered into the sequence an administrator should actually
work in. Each card carries severity and remediation effort, so you can
see what closes this afternoon with one config reload and what needs a
maintenance window. Severity alone gives you a report; severity with effort
gives you a work queue.
AI queue
Sessions ranked by how unusual they are, for a human to review first. This is the
only thing the AI layer does. Switch it off using the toggle in the header and
every score and finding stays identical — try it.
6 · Take it away
Export
JSON, HTML, PDF, and a CycloneDX 1.6 cryptographic bill of materials. All four
are views of one canonical result, so the number on screen and the number in the
PDF cannot disagree. The CBOM is the one to hand to a post-quantum migration
team — it lists what your servers were actually observed doing, not what a
configuration file claims.
What this tool will not tell you
It does not decrypt anything and holds no keys. It does not probe or contact any
server. It does not read message contents. And it will not call a downgrade an
attack unless the capture contains a signal that corroborates it.