# US 12,120,219 B2 · Machine learning-enabled character based encryption system > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,120,219 B2 - Document ID: US12120219B2 - Status: Granted - Kind: B2 - Application number: 17/987,425 - Filed: 2022-11-15 - Published: 2024-05-16 - Granted: 2024-10-15 - Earliest claimed: 2022-11-15 - Assignees: Sap SE - Applicants: Sap SE - Inventors: Ramalingam Tv - Web page: https://patentgrep.com/patent/US12120219B2 ## CPC - G09C 5/00: Ciphering apparatus or methods not provided for in the preceding groups, e.g. involving the concealment or deformation of graphic data such as designs, written or printed messages - H04L 9/0816: {Key establishment, i.e. cryptographic processes or cryptographic protocols whereby a shared secret becomes available to two or more parties, for subsequent use} - G06N 3/045: Combinations of networks - G06N 3/08: Learning methods - G06N 20/00: Machine learning - H04L 2209/08: Randomization, e.g. dummy operations or using noise - G09C: CIPHERING OR DECIPHERING APPARATUS FOR CRYPTOGRAPHIC OR OTHER PURPOSES INVOLVING THE NEED FOR SECRECY - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract In an example embodiment, a solution is provided for encrypting and decrypting data in which the solution itself creates unique symbols, reducing or eliminating the possibility that a hacker or other malicious actor can understand what the symbols mean, let alone decrypt them. More particularly, for encryption, the characters of an original text is split into individual vector-based characters, and each of these vector-based characters are split into subcharacters at intersection points. Each of the split characters are then rotated, and the rotated characters are organized one on top of each other. The characters are then connected to each other, and the intersection points of the subcharacters within the characters that were used to split the characters into subcharacters are passed into a machine learning model that is trained to add lines between intersection points that have no lines, as noise to further obfuscate the original text. ## Claims ### Claim 1 A system comprising: at least one hardware processor; and a computer-readable medium storing instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform operations comprising: accessing data, the data comprising one or more characters; identifying a language of the data, the language being read in a first direction; based on the language, retrieving a vector-based representation for each of the one or more characters; rotating each vector-based representation to produce a set of rotated vector-based representations; assigning an index value to each of the rotated vector-based representations, the index value based on a relative ordering of the one or more characters in the data for corresponding rotated vector-based representations; drawing the rotated vector-based representations in a single drawing so that borders of successive rotated vector-based representations in the relative ordering overlap each other, and such that the rotated vector-based representations are drawn in the relative ordering in a second direction perpendicular to the first direction; and passing the drawing into a machine learning model, the machine learning model adding additional vectors to the drawing as noise, producing an encrypted drawing. ### Claim 2 (depends on 1) The system of claim 1, further comprising assigning a color code to each of the rotated vector-based representations and to the noise, wherein the drawing the rotated vector-based representations includes drawing each rotated vector-based representation in a color corresponding to its assigned color code, wherein the machine learning model further adds the additional vectors to the drawing in a color corresponding to the color code assigned to the noise. ### Claim 3 (depends on 1) The system of claim 1, further comprising training the machine learning model by passing a labeled drawing having additional vectors added as noise to a machine learning algorithm, wherein a label in the labeled drawing is indicative of how difficult it is to determine the data from the labeled drawing. ### Claim 4 (depends on 3) The system of claim 3, wherein the labeled drawing is the encrypted drawing labeled by a user subsequent to the user being presented with the encrypted drawing in a graphical user interface. ### Claim 5 (depends on 3) The system of claim 3, wherein the labeled drawing is part of a set of labeled drawings passed to the machine learning algorithm during a dedicated training phase. ### Claim 6 (depends on 4) The system of claim 4, wherein the machine learning model is retrained by the machine learning algorithm using the labeled drawing. ### Claim 7 (depends on 1) The system of claim 1, wherein the direction of the rotating is dependent on the first direction. ### Claim 8 A method comprising: accessing data, the data comprising one or more characters; identifying a language of the data, the language being read in a first direction; based on the language, retrieving a vector-based representation for each of the one or more characters; rotating each vector-based representation to produce a set of rotated vector-based representations; assigning an index value to each of the rotated vector-based representations, the index value based on a relative ordering of the one or more characters in the data for corresponding rotated vector-based representations; drawing the rotated vector-based representations in a single drawing so that borders of successive rotated vector-based representations in the relative ordering overlap each other, and such that the rotated vector-based representations are drawn in the relative ordering in a second direction perpendicular to the first direction; and passing the drawing into a machine learning model, the machine learning model adding additional vectors to the drawing as noise, producing an encrypted drawing. ### Claim 9 (depends on 8) The method of claim 8, further comprising assigning a color code to each of the rotated vector-based representations and to the noise, wherein the drawing the rotated vector-based representations includes drawing each rotated vector-based representation in a color corresponding to its assigned color code, wherein the machine learning model further adds the additional vectors to the drawing in a color corresponding to the color code assigned to the noise. ### Claim 10 (depends on 8) The method of claim 8, further comprising training the machine learning model by passing a labeled drawing having additional vectors added as noise to a machine learning algorithm, wherein a label in the labeled drawing is indicative of how difficult it is to determine the data from the labeled drawing. ### Claim 11 (depends on 10) The method of claim 10, wherein the labeled drawing is the encrypted drawing labeled by a user subsequent to the user being presented with the encrypted drawing in a graphical user interface. ### Claim 12 (depends on 10) The method of claim 10, wherein the labeled drawing is part of a set of labeled drawings passed to the machine learning algorithm during a dedicated training phase. ### Claim 13 (depends on 11) The method of claim 11, wherein the machine learning model is retrained by the machine learning algorithm using the labeled drawing. ### Claim 14 (depends on 8) The method of claim 8, wherein the direction of the rotating is dependent on the first direction. ### Claim 15 A non-transitory machine-readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising: accessing data, the data comprising one or more characters; identifying a language of the data, the language being read in a first direction; based on the language, retrieving a vector-based representation for each of the one or more characters; rotating each vector-based representation to produce a set of rotated vector-based representations; assigning an index value to each of the rotated vector-based representations, the index value based on a relative ordering of the one or more characters in the data for corresponding rotated vector-based representations; drawing the rotated vector-based representations in a single drawing so that borders of successive rotated vector-based representations in the relative ordering overlap each other, and such that the rotated vector-based representations are drawn in the relative ordering in a second direction perpendicular to the first direction; and passing the drawing into a machine learning model, the machine learning model adding additional vectors to the drawing as noise, producing an encrypted drawing. ### Claim 16 (depends on 15) The non-transitory machine-readable medium of claim 15, further comprising assigning a color code to each of the rotated vector-based representations and to the noise, wherein the drawing the rotated vector-based representations includes drawing each rotated vector-based representation in a color corresponding to its assigned color code, wherein the machine learning model further adds the additional vectors to the drawing in a color corresponding to the color code assigned to the noise. ### Claim 17 (depends on 15) The non-transitory machine-readable medium of claim 15, further comprising training the machine learning model by passing a labeled drawing having additional vectors added as noise to a machine learning algorithm, wherein a label in the labeled drawing is indicative of how difficult it is to determine the data from the labeled drawing. ### Claim 18 (depends on 17) The non-transitory machine-readable medium of claim 17, wherein the labeled drawing is the encrypted drawing labeled by a user subsequent to the user being presented with the encrypted drawing in a graphical user interface. ### Claim 19 (depends on 17) The non-transitory machine-readable medium of claim 17, wherein the labeled drawing is part of a set of labeled drawings passed to the machine learning algorithm during a dedicated training phase. ### Claim 20 (depends on 18) The non-transitory machine-readable medium of claim 18, wherein the machine learning model is retrained by the machine learning algorithm using the labeled drawing.