# US 2018/0322279 A1 · Providing differentially private data with causality preservation

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2018/0322279 A1
- Document ID: US20180322279A1
- Status: Published application
- Kind: A1
- Application number: 15/701,826
- Filed: 2017-09-12
- Published: 2018-11-08
- Earliest claimed: 2017-05-02
- Assignees: Sap SE
- Applicants: Sap SE
- Inventors: Wasilij Beskorovajnov; Daniel Bernau
- Web page: https://patentgrep.com/patent/US20180322279A1

## CPC

- G06F 21/552: {involving long-term monitoring or reporting}
- G06F 16/215: Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
- G06F 16/221: {Column-oriented storage; Management thereof}
- G06F 21/6245: {Protecting personal data, e.g. for financial or medical purposes}
- G06F 21/6254: {by anonymising data, e.g. decorrelating personal data from the owner's identification}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method is disclosed for providing sanitized log data to a threat detection system. The sanitized log data is derived from a log table with continuous columns, themselves having continuous entries with continuous values. First, a retention probability parameter and an accuracy radius parameter are selected. Next, a probability distribution function is initialized with the retention probability parameter and the accuracy radius parameter. For each continuous value, the probability distribution function is applied, resulting in perturbed continuous values of a perturbed continuous columns Finally, the perturbed continuous columns are provided as the sanitized log.

## Claims

### Claim 1

A method for providing sanitized log data to a threat detection system, the sanitized log data being derived from a log table, the log table having at least one continuous column having continuous entries with continuous values, the method comprising:
selecting a retention probability parameter and an accuracy radius parameter; initializing a probability distribution function with the retention probability parameter and the accuracy radius parameter; and
for each continuous value in the continuous values applying the probability distribution function to the continuous value to compute a perturbed continuous value, resulting in at least one perturbed continuous column; and
providing the at least one perturbed continuous column as the sanitized log.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
rounding the perturbed continuous value after adding floating-point Laplacian noise associated with the probability distribution function to the continuous value.

### Claim 3 (depends on 1)

The method of claim 1, wherein the log table has at least one discrete column in addition to the at least one continuous column, the discrete column having discrete entries with discrete values, the method further comprising:
choosing a randomization parameter;
specifying a domain for possible discrete values;
for each discrete value in the discrete values, determining whether to keep the original discrete value or randomly selecting a value from the domain of possible values.

### Claim 4 (depends on 3)

The method of claim 3, further comprising:
de-identifying the entries with discrete values.

### Claim 5

A threat detection system comprising:
an intrusion detection system sensor for providing sanitized log data derived from a log table to the threat detection system, the log table having at least one continuous column having continuous entries with continuous values; and
a privacy module for selecting a retention probability parameter and an accuracy radius parameter, wherein the privacy module initializes a probability distribution function with the retention probability parameter and the accuracy radius parameter, wherein for each continuous value in the continuous values the probability distribution function is applied to the continuous value to compute a perturbed continuous value, resulting in at least one perturbed continuous column, and wherein the at least one perturbed continuous column is provided to an intrusion detection system repository that by way of queries and responses provides the sanitized log to at least one data analyst.

### Claim 6 (depends on 5)

The threat detection system of claim 5, wherein the privacy module rounds the perturbed continuous value after adding floating-point Laplacian noise associated with the probability distribution function to the continuous value.

### Claim 7 (depends on 5)

The threat detection system of claim 5, wherein the log table has at least one discrete column in addition to the at least one continuous column, the discrete column having discrete entries with discrete values, wherein the privacy module chooses a randomization parameter and specifies a domain for possible discrete values, and wherein for each discrete value in the discrete values the privacy module determines whether to keep the original discrete value or to randomly select a value from the domain of possible values.

### Claim 8 (depends on 7)

The threat detection system of claim 7, wherein the entries with discrete values are de-identified.

### Claim 9 (depends on 8)

The threat detection system of claim 8, wherein the entries with discrete values are de-identified before the value is randomly selected.

### Claim 10

An enterprise application landscape comprising:
at least one enterprise software system;
an enterprise threat detection system for providing at least one data analyst with sanitized log data derived from a log table associated with the enterprise software system, the log table having at least one continuous column having continuous entries with continuous values; and
a privacy module for selecting a retention probability parameter and an accuracy radius parameter, wherein the privacy module initializes a probability distribution function with the retention probability parameter and the accuracy radius parameter, wherein for each continuous value in the continuous values the probability distribution function is applied to the continuous value to compute a perturbed continuous value, resulting in at least one perturbed continuous column, and wherein the at least one perturbed continuous column is provided to an intrusion detection system repository that by way of queries and responses provides the sanitized log to the data analyst.

### Claim 11 (depends on 10)

The enterprise application landscape of claim 10, wherein the privacy module rounds the perturbed continuous value after adding floating-point Laplacian noise associated with the probability distribution function to the continuous value.

### Claim 12 (depends on 10)

The enterprise application landscape of claim 10, wherein the log table has at least one discrete column in addition to the at least one continuous column, the discrete column having discrete entries with discrete values, wherein the privacy module chooses a randomization parameter and specifies a domain for possible discrete values, wherein for each discrete value in the discrete values the privacy module determines whether to keep the original discrete value or to randomly select a value from the domain of possible values.

### Claim 13 (depends on 12)

The enterprise application landscape of claim 12, wherein the entries with discrete values are de-identified.

### Claim 14 (depends on 13)

The enterprise application landscape of claim 13, wherein the entries with discrete values are de-identified before the value is randomly selected.
