Except that security standards are mostly driven by compatibility with all sorts of edge cases (ancient versions of software! COBOL-bound CAs! tiny smart cards! huge distributed systems!) that are probably not relevant to your case and create big random security holes.
If you have a reasonable way of distributing software updates, signed-DH with an AES-based AEAD with termination detection for transport authentication, and a signed hash tree for data authentication, is going to be more secure than RANDOM_SECURITY_STANDARD.
Thankfully, there are some good well-written implementations of that (https://www.tarsnap.com/spiped.html, https://nacl.cr.yp.to), but picking a standard randomly will not get you them.
"You are astoundingly unlikely to create something better and more-secure from seeing a blog-post, reading some Wikipedia articles, etc. than a team of PhDs who did nothing but study mathematical theory, cryptography, and code-breaking for the last few years, and accordingly are vanishingly unlikely to create a new scheme which is more secure than existing standards once you account for performance and maintenance."