What Sanpu17 covers and what it does not
Three defined areas, clear boundaries and completely open content — no sign-up and no payment. Here is exactly what each one gives you, who it suits, and what you will not find here.
This page describes a blog, not a consultancy. There is no quotation here, no engagement and no consultant–client relationship. What there is: three defined content areas, and in each one I set out what I cover, which sources I rely on and what a reader will not find from me. The point is to save you time — if your topic is here, I have probably written about it already; if it is not, you will know within the first minute.
What you get: technical prose in English, open without sign-up and without payment, with tested code, public sources and an update date on every article. What you do not get: promises, paid mentions and conclusions you cannot check.
| Area | Who it is for | What you get | What it is not |
|---|---|---|---|
| Security architecture analysis | Architects and security engineers in a financial institution | Trust, authorisation and encryption models set against public standards | Not an audit-readiness review and not an opinion on a product |
| Trading infrastructure reviews | Systems, operations and development engineers at exchanges and payment processors | The order path, matching engines and protocols, on the basis of public documentation | Not investment advice and not a trading strategy |
| Applied cryptography guides | Developers implementing encryption, keys or TLS | Working code, sample configuration and a key hierarchy | Not a penetration test and not a sign-off that your implementation is secure |
Security architecture analysis
The analyses run between two end points. On one side the protocol: what goes on the wire, who authenticates whom, what a token looks like, and where the chain of trust between two services breaks. On the other side the infrastructure: where the keys run, who can read them, and what the hierarchy between the KMS and the HSM looks like. Between the two sits what actually interests me: the authorisation model — who may do what, on which object, and under which condition.
Most of the failures I come across are not weak encryption but the join between two components that each work as designed, with a wrong assumption on one side about the other. So every analysis starts with the question of what is assumed here, and what happens when the assumption blows up.
The standards I rely on
My comparison frame is built from open documents. NIST CSF serves as a map of categories — identify, protect, detect, respond and recover — not as a task list. OWASP, in particular OWASP Top Ten and OWASP API Security Top 10 (2023 edition) is the starting point for attack vectors; BOLA, the first category there, is often the whole article. PCI DSS 4.0.1 is the current version of the standard, and I treat it as a layer of requirements, not as proof that anything is secure. On Zero Trust I lean on NIST SP 800-207 from 2020, which defines the model without selling anything.
Length, depth and who it suits
The articles are long: between 14 and 21 min read each, with diagrams and code that actually runs. That is deliberate. A topic like mTLS between services does not end with a paragraph about certificates; it ends when you understand what happens when the certificate expires while the pod is restarting, and why that happens at peak hour. Every claim arrives with the place where you can check it: an RFC, a standards document, or the behaviour of an open-source tool.
This suits architects, security engineers and platform teams in a regulated institution. It does not suit anyone looking for a checklist to hand to an auditor, a yes-or-no answer on meeting a requirement, or a comparison that will tell them which product to buy. I do not do that, not because I have no opinion but because I do not have your data.
No audit opinion, no statement of compliance with a standard, and no review of your architecture under NDA. If you need someone to sign a document, that is the job of an accredited party, not of a blog. Nor will you get vendor rankings: I do not run these products in production.
Trading infrastructure reviews
Trading infrastructure is the world where security and real time collide. The reviews focus on two layers: the exchange and the clearing side. On the exchange side I take apart the order path, from the moment an order arrives at the gateway, passes risk checks, reaches the matching engine and is recorded in the order book. On the clearing side I deal with settlement: what is finalised in real time, what is batched, and where a gap opens between what the system knows and what actually happened.
One recurring question is FIX versus streaming. FIX remains the standard for connecting clients, because it is what the other side already speaks. But the internal path of an exchange has, in most places, moved to binary formats and streaming. In the articles I show where that boundary runs and why an architecture that mixes the two without distinction produces failures that are hard to diagnose. To understand the data pipeline I rely on the public documentation of Kafka and of Flink.
On latency I write in terms of distribution rather than average: the tail is what decides at the opening auction. For the same reason I check what happens when the order is late, when a message arrives twice, and when the matching engine replied and the client did not receive it.
The boundary: every review is based on public documentation — exchange specifications, protocol documents and open papers. There is no inside information here and no client data. And in one line: no investment recommendations and no trading strategies. What may be traded and when to buy — not on this site, in any wording.
Applied cryptography guides
This part was born out of frustration. There is plenty of material on what AES is and too little on what you do when you need to rotate a key without stopping the environment. So the guides deal with implementation: how to encrypt data at rest and traffic in transit, how to build a key hierarchy with KEK and DEK, and what key rotation looks like when it does not drop requests halfway through.
On HSMs I cover what affects design: the FIPS 140-3 security levels from 1 to 4, where level 3 is the practical bar for most financial infrastructure — tamper resistance and identity-based authentication. On TLS I stay with RFC 8446 (TLS 1.3) and leave behind what came before it, and explain what happens when certificate-based client authentication in the style of RFC 8705 is attached to it. Wherever PCI DSS touches key management I cite it as context, not as a claim that your implementation meets the requirement.
The code in the guides is tested: I run every example at least once in the lab before publishing, including the failure steps, because an example that fails silently is worse than no example. The configurations come from tools you can run locally, not from a management interface you cannot reproduce.
And what it is not: not a penetration test, and not a sign-off that your implementation is secure. A good guide teaches you to ask the right questions about your own system; it does not replace someone looking at it. If after reading you know what to check, that is what I intended.
What the writing process looks like
Every article goes through the same path. The order is what separates a technical article from a summary of vendor slides.
- Research. Starts with RFCs and standards documents — NIST, PCI SSC, OWASP — continues to the tool documentation, and ends in the lab: running the thing I intend to write about.
- Draft. First the argument, then the diagram, then the code. Every claim gets a source, and a claim without a source is marked as my own judgement.
- Technical review. Someone working in the field reads the draft and looks for what does not hold: wrong assumptions, an outdated version, an example that would not have passed in production.
- Publication. With a date, the author’s name and a list of sources you can open. No internal sources and no references to closed documents.
- Update. When a version of a standard changes, an error surfaces or a tool’s behaviour changes, the article is updated and the date changes. There is no silent editing: a substantive change gets a note in the body of the text.
A year-old article with a current update date is worth more than a new one without a date. I would rather update an existing article than publish a second version.
Found a mistake? Write to me, and that is enough. I check the source, fix it, and add a correction line with the date. If the mistake was substantive — a conclusion resting on a wrong assumption, for instance — the note appears at the top of the article, not at the bottom. Fast correction is what keeps trust over time.
What Sanpu17 does not cover
The list of exclusions is definitive, not a disclaimer. If one of these items is what you are looking for, this site will not help.
- Not investment advice — no recommendations on assets, no pricing and no trading strategies.
- Not penetration testing — no vulnerability scans, no experiments in your environment and no findings reports.
- Not a SOC and not monitoring — no 24/7 operations centre, no incident response and no intrusion investigation.
- Not standards compliance audits — no PCI DSS audit, no ISO 27001 and no opinion for a regulator.
- Not legal advice — no binding interpretation of a law or regulation. The MAS Technology Risk Management Guidelines appear here as context only, not as guidance.
- Not products — no shop, no integrations and no implementation work.
- Not sponsored content — no paid posts, no vendors paying for a mention and no rankings that were bought.
- Not training — no courses, no certification and no workshops. The material is open, and that is all it is meant to be.
Who it is for
Three kinds of reader get more out of this than others.
- Architects and security engineers designing service identity, authorisation and encryption, who need an explanation at the level of the protocol rather than the level of the slide.
- Developers on payments and trading teams writing code that touches keys, tokens and requests, who want to see an implementation before it reaches production.
- Team leads and risk and compliance staff who need to understand what is required and what is not, so they can hold a conversation with engineering without translating every term.
If you need an opinion on an existing system, a test or hands-on support, that is not here.
How to start
The simplest starting point is the article archive: six articles with a summary and a date, by the topic on your mind right now. If a term stops you, the Glossary gives a short definition and a link to go deeper.
For anyone starting from scratch I recommend the article on Zero Trust in financial systems. It touches identity, mTLS and policy enforcement, and all three recur in every other topic, so reading it shortens the way to the rest.
A question, a mistake or a topic suggestion? You can reach me through the contact page. I read every message and answer technical questions, but I do not give specific advice on your system, even when the question is well put.