securedna

12 atoms · 跨 1 天 · 首见 2026-06-04 · 最近 2026-06-04

三色: 🟦 fact 11 · 🟥 take 1 · stance ▲0/▼0/◆12

来源: X 12

时态: fresh:12

标签: 好观点:11 · 好信源:1

叙事 (narrative) (10)

• 2026-06-04 · design approach
  starting out, you want to screen for ~20-30mer fragments of hazardous organisms. why screen? https://t.co/ryMxv1eEhP
  tags: 好观点 · → daily
• 2026-06-04 · screening design
  in your first design, just take plaintext FASTA files from providers, shingle them into all 20-30mer windows (plus reverse complement, etc) and check if they're in your pool of hazardous fragments.
  tags: 好观点 · → daily
• 2026-06-04 · partner strategy
  you partner with as many providers (and benchtop manufacturers) as possible so you can correlate customer orders across providers to catch order splitting, and to act as an central, private repository for newly-discovered, not-yet-published hazards.
  tags: 好观点 · → daily
• 2026-06-04 · variant detection approach
  to your hazard pool, you add predicted functional variants based on a variant generation pipeline.
  tags: 好观点 · → daily
• 2026-06-04 · flagging heuristics
  tune the flagging heuristics based on window hit rate. (e.g. by adding 'reverse screening' - finding which windows show up a lot in non-hazardous orders)
  tags: 好观点 · → daily
• 2026-06-04 · privacy design flaw
  this system you've built has TERRIBLE privacy properties! you've introduced yourself as a central handler of valuable customer sequences, and a database of recently-discovered, nonpublic hazardous organisms for malicious actors to steal.
  tags: 好观点 · → daily
• 2026-06-04 · hashing approach
  hash the fragments: the database holds hashes, and customers send you hashes. this limits what you can do at runtime - you can only do exact matches, no distance - but now anyone who steals a database just gets opaque hashes
  tags: 好观点 · → daily
• 2026-06-04 · DOPRF approach
  we somehow need a rate-limited hash function... which distributed OPRFs give us! DOPRFs let us shard the hash function over multiple servers, so users need multiple network calls to perform the hash function, while still blinding the original input from those servers.
  tags: 好观点 · → daily
• 2026-06-04 · rate limiting design
  by using DOPRFs, our keyservers (the servers that hold a shard of the DOPRF master key and offer the distributed hashing) can detect when a single user certificate is requesting too many hashes, and globally yank their token. that way, even if an attacker has physical possession of database files, they're completely useless!
  tags: 好观点 · → daily
• 2026-06-04 · certificate system
  the final layer is a custom certificate system that lets legitimate researchers easily get permission to work with organisms relevant to their research: instead of juggling stacks of paper forms and emails, they simply get issued a signed token from their university or other relevant organization authorizing them to ignore hits on certain organism ranges
  tags: 好观点 · → daily

事实 (fact) (2)

• 2026-06-04 · status
  a free to use, swiss nonprofit screening organization supported by researchers and used by synthesis providers in the US, EU, and China. :-)
  tags: 好信源 · → daily
• 2026-06-04 · screening length rationale
  in wet lab experiments that's about the length of fragments that can be stuck back together into the original sequence reliably.
  tags: 好观点 · → daily

时间轴 (近 20)

• 2026-06-04 · narrative · design approach · → daily
• 2026-06-04 · fact · screening length rationale · → daily
• 2026-06-04 · narrative · screening design · → daily
• 2026-06-04 · narrative · partner strategy · → daily
• 2026-06-04 · narrative · variant detection approach · → daily
• 2026-06-04 · narrative · flagging heuristics · → daily
• 2026-06-04 · narrative · privacy design flaw · → daily
• 2026-06-04 · narrative · hashing approach · → daily
• 2026-06-04 · narrative · DOPRF approach · → daily
• 2026-06-04 · narrative · rate limiting design · → daily
• 2026-06-04 · narrative · certificate system · → daily
• 2026-06-04 · fact · status · → daily

← 实体目录 · 系统日志