Get the newsletter
LG · Security · Filed Jan 6, 2026 · USPTO published Nov 19, 2026 · This week's release

LG Files Patent for Faster Computing on Data That Stays Encrypted

Doing math on locked data is so slow that one example in LG's filing balloons to 4 to the 365th power possible routes through a single program. LG's patent describes a way to throw almost all of them out before the work starts.

A privacy setup scrambles a person's face for protection while letting a remote computer analyze their body normally. Drawing from patent filing US 2026/0370448 A1.
A privacy setup scrambles a person's face for protection while letting a remote computer analyze their body normally.All 16 drawings ↓
Publication number
US 2026/0370448 A1(US20260370448A1)
Applicant
LG ELECTRONICS INC.
Filed · Published
Jan 6, 2026 · Nov 19, 2026
Inventors
Hajoon KO, Jaeky OH, Byoungwoo YOON, Minwook JEONG, Sunim LEE, Jongho SHIN
Status when we published
Waiting for an examiner
Document
20 claims
US classification
380/28

What LG's locked-data shortcut actually does

You're a small business owner uploading a year of daily sales and expenses to a cloud service that will total them up. You'd like the service to do the math without ever seeing a single figure. A special kind of encryption makes that possible, and it has a big downside: it can run thousands of times slower than ordinary computing.

Much of the slowdown comes from if-this-then-that decisions. When the server can't see the locked numbers, it can't tell which way a decision goes, so it works through every possible route. Over 365 days, that adds up to an absurd pile of routes.

LG's patent lets you mark some inputs as fine to share, like which days you want counted. The server uses those open inputs to rule out nearly every route up front and does the encrypted math on just the one that applies. Your dollar amounts stay locked.

From the filing · CLAIM 1
… classify variables in the source code as plaintext input variables or ciphertext input variables; generate code fragments corresponding to execution paths in the source code, excluding execution paths in which branch conditional statements include the variables classified as ciphertext input variables …

Translation: The system sorts program variables into unencrypted and encrypted types, then skips code paths where encrypted data controls decisions.

How LG sorts locked inputs from shareable ones

The system starts with the source code of an ordinary program and sorts its inputs into ciphertext (encrypted, hidden from the server) and plaintext (readable by the server). A user can make that choice on screen.

Next it runs dynamic symbolic execution, a testing technique that runs the program with stand-in values to map every route, or execution path, through the code. Any decision that depends only on plaintext inputs gets settled ahead of time: each route becomes its own code fragment with that decision stripped out. A standard encryption-aware compiler turns the fragments into homomorphic encryption binary fragments (ready-to-run pieces that compute on locked data). The decisions themselves are stored separately and never compiled.

When a request arrives, the system checks the plaintext inputs against the stored decisions and runs only the matching fragment. The filing also covers some edge cases:

  • If a decision depends on a locked input, the system prunes what it can before that point and keeps the rest as ordinary multi-route code.
  • If a math solver can't work out a decision in time, the user can be asked to keep searching or type in a solution.
  • A screen can show a computational load reduction rate. In the filing's example, freeing one input shows 12%, and encrypting everything shows 0%.

In the filing's bookkeeping example, 4 to the 365th power routes shrink to one.

From the filing · THE ABSTRACT
… compile the generated code fragments to obtain homomorphic encryption binary fragments.

Translation: It turns the processed pieces of code into binary files that can run calculations on encrypted data without ever decrypting it.

Why faster encrypted math matters for cloud privacy

If encrypted services get faster without giving up privacy, more of them can run in the cloud instead of on your device. The filing points to smart home gadgets and home health services, where sensitive readings leave your house and you'd rather the server never see them. Faster encrypted processing also means less waiting for you and less strain on the company's servers.

There is a tradeoff. The server learns whatever inputs get marked as shareable, such as which days were selected in the example, so the savings depend on how much can be safely opened up. The filing describes a developer tool that shows that tradeoff with a readout of the computing saved. It does not tie the idea to a specific LG product.

LG's third filing we've tracked since November in our LG coverage follows the flip-text see-through screen and the anti-haze foldable cover.

Source: US 2026/0370448 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time LG patent news breaks.

There are more where this came from

We read the new patents from Big Tech, carmakers and more and send you the ones worth knowing. Plain English, free, every week.

The drawings · 16 sheets

16 drawing sheets from US 2026/0370448 A1 · click any drawing to enlarge

Cloud setups that expose raw data during calculation, drawing sheet 1 from US 2026/0370448 A1
Sheet 1 Cloud setups that expose raw data during calculation
Keeping customer data encrypted while an AI model runs, drawing sheet 2 from US 2026/0370448 A1
Sheet 2 Keeping customer data encrypted while an AI model runs
Splitting private faces from ordinary data to save computing effort, drawing sheet 3 from US 2026/0370448 A1
Sheet 3 Splitting private faces from ordinary data to save computing effort
Sample software rules that lead down different calculation paths, drawing sheet 4 from US 2026/0370448 A1
Sheet 4 Sample software rules that lead down different calculation paths
Mapping every potential decision path through a computer program, drawing sheet 5 from US 2026/0370448 A1
Sheet 5 Mapping every potential decision path through a computer program
Exploring every possible branch when all inputs stay encrypted, drawing sheet 6 from US 2026/0370448 A1
Sheet 6 Exploring every possible branch when all inputs stay encrypted
Skipping unused paths when noncritical inputs are unencrypted, drawing sheet 7 from US 2026/0370448 A1
Sheet 7 Skipping unused paths when noncritical inputs are unencrypted
Preparing prebuilt code paths to run encrypted calculations quickly, drawing sheet 8 from US 2026/0370448 A1
Sheet 8 Preparing prebuilt code paths to run encrypted calculations quickly
Show the other 8 sheets
Steps for analyzing code and running computations on encrypted data, drawing sheet 9 from US 2026/0370448 A1
Sheet 9 Steps for analyzing code and running computations on encrypted data
An interface where programmers select which inputs require encryption, drawing sheet 10 from US 2026/0370448 A1
Sheet 10 An interface where programmers select which inputs require encryption
Zero computational savings when every input remains fully encrypted, drawing sheet 11 from US 2026/0370448 A1
Sheet 11 Zero computational savings when every input remains fully encrypted
Estimated computing gains after leaving some data unencrypted, drawing sheet 12 from US 2026/0370448 A1
Sheet 12 Estimated computing gains after leaving some data unencrypted
Asking permission to analyze broken code more deeply, drawing sheet 13 from US 2026/0370448 A1
Sheet 13 Asking permission to analyze broken code more deeply
Requesting manual input to resolve an unsupported code branch, drawing sheet 14 from US 2026/0370448 A1
Sheet 14 Requesting manual input to resolve an unsupported code branch
Connected chips and memory units powering the secure computing setup, drawing sheet 15 from US 2026/0370448 A1
Sheet 15 Connected chips and memory units powering the secure computing setup
Hardware inside the user device and cloud server, drawing sheet 16 from US 2026/0370448 A1
Sheet 16 Hardware inside the user device and cloud server
Patent filing page

Watchlists this filing belongs to

All watchlists →
Reader comments

No comments yet

Start the discussion

Real name or a handle, either is fine. Comments are read by a person before they appear, so allow a little time. Keep it about the filing.