# US 2020/0404007 A1 · Systems and Methods for Detecting Injection Exploits

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

- Number: US 2020/0404007 A1
- Document ID: US20200404007A1
- Status: Published application
- Kind: A1
- Application number: 16/844,915
- Filed: 2020-04-09
- Published: 2020-12-24
- Earliest claimed: 2019-04-09
- Assignees: Prismo Systems Inc.
- Applicants: Prismo Systems Inc.
- Inventors: Abhishek Singh; Ramesh Mani; Anjan Venkatramani; Chihwei Chao
- Web page: https://patentgrep.com/patent/US20200404007A1

## CPC

- H04L 63/1416: {Event detection, e.g. attack signature detection}
- G06F 16/24537: {of operators}
- G06F 21/552: {involving long-term monitoring or reporting}
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods for detecting injection exploits in a networked computing environment are disclosed. In one embodiment, a method for detection includes monitoring web applications that are executing and detecting when an execution function is received over a network and invoked, where an execution function is a function that accepts external free-form data values, detecting malicious code by generating a model of legitimate behavior subsequent to invocation of the execution function, comparing actual behavior to the model of legitimate behavior, and generating an alert when the actual behavior deviates from the model of legitimate behavior and validating whether the deviation of the actual behavior is due to one or more functions that accept external input.

## Claims

### Claim 1

A method for detecting injection exploits in a networked computing environment, the method comprising:
using a collector server, monitoring web applications that are executing and detecting when an execution function is received over a network and invoked, where an execution function is a function that accepts external free-form data values;
detecting malicious code by:
generating a model of legitimate behavior subsequent to invocation of the execution function;
comparing actual behavior to the model of legitimate behavior; and
generating an alert when the actual behavior deviates from the model of legitimate behavior and validating whether the deviation of the actual behavior is due to one or more functions that accept external input, by a collector server.

### Claim 2 (depends on 1)

The method of claim 1, where the execution function is a program execution function; and
wherein comparing actual behavior to the model of legitimate behavior comprises generating a dynamic call graph and identifying a new or deleted code path in the actual behavior that is not present in the model, and determining whether the new code path is due to external input.

### Claim 3 (depends on 1)

The method of claim 1, where the execution function is a program execution function and wherein comparing actual behavior to the model of legitimate behavior comprises generating and identifying a new or deleted node in the AST that is not present in the original model and validating whether the new or deleted node is due to external input.

### Claim 4 (depends on 1)

The method of claim 1, where the execution function is a deserialization function;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises generating a model of expected application programming interface (API) calls; and
wherein comparing actual behavior to the model of legitimate behavior comprises identifying API calls in the actual behavior that are not in the model of expected API calls.

### Claim 5 (depends on 1)

The method of claim 1, wherein the execution function is a database query comprising a command and arguments;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises generating an abstract syntax tree (AST) by extracting and normalizing the arguments of the SQL query; and
wherein comparing actual behavior to the model of legitimate behavior comprises identifying an additional node in the AST and the cause of the additional node.

### Claim 6 (depends on 1)

The method of claim 1, wherein the execution function is an LDAP filter query comprising a command and arguments;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises generating an abstract syntax tree (AST) by extracting and normalizing the arguments of the SQL query; and
wherein comparing actual behavior to the model of legitimate behavior comprises identifying an additional node in the AST and the cause of the additional node.

### Claim 7 (depends on 1)

The method of claim 1, wherein the execution function is an LDAP filter query comprising a command and arguments;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises checking for HTML elements in the arguments of the execution functions which is not present in the original model; and
comparing if the HTML elements are from external inputs.

### Claim 8 (depends on 1)

The method of claim 1, wherein the execution function is a database query comprising a command and arguments;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises checking for HTML elements in the arguments of the execution functions which is not present in the original model; and
comparing if the HTML elements are from external inputs.

### Claim 9 (depends on 1)

The method of claim 1, wherein the execution function is an API command that accepts a URI as an input parameter and opens a network object denoted by the URI;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises generating a program dependency graph; and
wherein comparing actual behavior to the model of legitimate behavior comprises identifying APIs which make network connections and which accept user inputs, determining the destination type of the IP address in the network connection, and determining whether the parameters passed to make the network connection are from a method which accepts external input.

### Claim 10 (depends on 1)

The method of claim 1, wherein the execution function is a function that passes data to stored procedures of a database;
wherein generating a model of legitimate behavior subsequent to invocation of the execution function comprises creating an abstract syntax tree (AST) using one or more honey queries;
wherein comparing actual behavior to the model of legitimate behavior comprises comparing the AST of a normalized honey query with the AST of a honey query having arguments passed to stored procedures.

### Claim 11 (depends on 1)

The method of claim 1, wherein generating an alert when the actual behavior deviates from the model of legitimate behavior further comprises performing a remediation action.

### Claim 12 (depends on 9)

The method of claim 9, wherein the remediation action is generating an indicator of compromise (IoC) for a suspected malicious IP address.

### Claim 13 (depends on 9)

The method of claim 9, wherein the remediation action is blocking further API calls by the execution function.

### Claim 14 (depends on 9)

The method of claim 9, wherein the remediation action is sanitizing the arguments passed to the execution function and allowing the execution function to execute.

### Claim 15 (depends on 1)

The method of claim 1, further comprising generating an entity risk score for an entity on the network and displaying the entity risk score to a user, where the entity risk score is computed from a number of detected incidents on the entity, a decay factor over time, and an importance factor of the entity.

### Claim 16

A method for linking events generated by a web application agent and an operating system monitoring agent, the method comprising:
capturing a web application transaction by creating a graph that originates at an entry point event identifier (ID) using an application agent on a sensor server;
invoking a system process at the exit point and capturing the process identifier (PID) of the processes that are spawned as a result using the application agent on the sensor server;
sending a process event comprising the PID and transaction data about the web application transaction from the sensor server to a collector server;
querying for the current UUID created by the process monitor for the PID when the process event is received by the collector server;
applying the retrieved UUID for the subprocess as a new attribute to the entry point event for the web application transaction; and
querying for entry point events having the same UUID to continue tracing into web application transactions.

### Claim 17

A method for correlating transaction components of a transaction that calls a different process without header decoration, the method comprising:
identifying and tracking a publisher side of a transaction in an electronic publisher-subscriber system having an asynchronous external trace decoration mechanism;
assigning a unique identifier (ID) and sending trace information of the publisher side of the transaction to a collector system and cache the ID;
sending the unique ID and a message ID obtained by a publisher to the collector system when an application callback is invoked with the message ID;
identifying and tracing a subscriber side of the transaction and extracting the message ID;
sending trace information of the subscriber side of the transaction with the message ID to the collector system; and
correlating the trace information of the publisher side of the transaction and the trace information of the subscriber side of the transaction using the message ID.
