Microsoft Technology Licensing, LLC

Roshanak Zilouchian Moghaddam

Neelakantan Sundaresan

Yevhen Mohylevskyy

Jason Shaver

Spandan Garg

Kalpathy Sitaraman Sivaraman

Aaron Yue-chiu Chan

Colin Bruce Clement

Anant Girish Kharkar

Paul Sean Harrington

Chen Wu

Matthew Glenn Jin

Shaun Miller

Yijin Wei

Christopher O'Toole

Shengyu Fu

Xiaoyu Liu

Ying Zhao

Eslam K Abdelreheem

Xin Shi

Gearard Boland

Siyu Yang

Jorge Banuelos

Payal Bajaj

Anshul Gupta

Eduardo Enrique Noriega De Armas

Kyle P Jackson

Alexander Winthrop Helyar

Ahmed Ezzat Abdelgawad Magooda

Mohamed Ali Mohamed Elkamhawy

An example generator tool generates an example illustrating correct usage of a command of a command line interface. A command may include a command name, zero or more subcommands, and one or more parameters with a corresponding parameter value. A template containing the correct syntax of the command is obtained from a template database. Parameter values for the template are generated from a neural transformer with attention given the command template.

Neural command line interface example generation

A code completion system for a CLI utilizes neural transformer models with attention to generate candidates to complete a line of CLI code. The code completion system uses a first deep learning model to predict at most k candidate command names to follow n immediately preceding lines of CLI code which are presented to a developer. Upon the developer accepting one of the candidate command names, the code completion system uses a second deep learning model to predict at most k parameter strings to complete the line of CLI code.

Neural transformer code completion for command line interface

A false positive vulnerability system detects whether a software vulnerability identified by a static code vulnerability analyzer is a true vulnerability or a false positive. The system utilizes deep learning models to predict whether an identified vulnerability is accurate given the source code context of the identified vulnerability. A neural encoder transformer model is trained to classify a false positive given the method body including the identified vulnerability. A neural decoder transformer model is trained to predict a candidate line-of-code to complete a prompt inserted into the context of the identified vulnerability. The candidate line-of-code that successfully completes the prompt is used as a signal to identify that the identified vulnerability is a false positive.

False positive vulnerability detection using neural transformers

A vulnerability detection and repair system utilize a classifier model to detect a software vulnerability in a source code snippet and the tokens in the source code snippet attributable to the vulnerability. A large language model is then given the vulnerable source code snippet, its vulnerability type, the vulnerability tokens, and a few-shot examples to determine whether or not the source code snippet includes the identified vulnerability. The few-shot examples include positive and negative samples of the type of vulnerability to guide the large language model towards the correct output.

Localizing vulnerabilities in source code at a token-level

A source code repair system generates a prompt including a few-shot examples of code changes made previously to correct a particular source code vulnerability. The prompt is given to a large language model to generate repair code for a source code snippet having the same vulnerability. Code changes made for a particular source code vulnerability, in the form of code diffs, are clustered into groups of closely related code change embeddings. A select few of the code diffs in each group having a closest mean of the cluster are used as the few-shot examples for the corresponding source code vulnerability.

Code-change based prompt for code repair

Example solutions performing software code vulnerability reduction. An input code portion is extracted from input software code. The input code portion may be syntactically incomplete and/or syntactically incorrect. A code vulnerability is detected in the input code portion. A correction of the code vulnerability is made, and an output code portion is generated including the correction. In some examples, a code vulnerability detection tool takes, as input, the output from a code completion tool. The output is thus annotated or corrected in real-time, as a user is developing the code.

Vulnerability reduction for syntactically incomplete code

A pre-trained neural code generation model generates repair code for a method containing a performance bug given a prompt including a code transformation instruction. The code transformation instruction guides the model on how to predict the repair code when the model has not been fine-tuned for the repair code task. The code transformation instruction is retrieved from abstract bug patterns derived from historical performance bug fixes found in commits to a source code repository. The augmentation of the code transformation instruction in the prompt to the pre-trained neural code generation model provides the model with a hint on how the repair code may be generated based on similar performance bug fixes.

Performance bug repair via retrieval-augmented neural code generation model

A failure recommendation system for a command line interface (CLI) uses machine learning to predict the most likely command to correct an unsuccessful or failed attempt to perform an intended operation using the CLI. The failure recommendation system is based on a conditional probability model trained on failure-success pairs of commands from CLI telemetry data to learn the most likely command to remediate a failure. The conditional probability model predicts the most likely command based on a failure type and the failed command. The failure type is identified through a failure type classifier and is used to select the most likely command to remediate a failure from the different events that may lead to a failure.

Failure recovery recommendations for CLI commands

A computer implemented method includes accessing performance trace data for executed code of multiple services. Symbols corresponding to functions of the executed code are identified. First sequences of functions from the identified symbols are identified and a first performance threshold for each identified first sequence of functions is computed. The method includes receiving an incoming performance trace, detecting second sequences of functions from the incoming performance trace, identifying second sequences equivalent to the first sequences, and comparing performance of the identified second sequences to the first performance threshold for each of the equivalent first sequences to identify second sequences as comprising a performance bottleneck.

System and method for identifying performance bottlenecks

A neural classifier model is used to detect cybersecurity vulnerabilities in the source code predicted by a deep learning code generation model having been trained on source code possibly containing security bugs. Upon the classifier model classifying a given source code snippet as likely containing a cybersecurity vulnerability, a proposed repair for the cybersecurity vulnerability is predicted from a neural decoder transformer model having been trained on non-vulnerable source code. The neural decoder transformer model is used to predict source code that repairs the cybersecurity vulnerability given the source code classified with a cybersecurity vulnerability.

Source code vulnerability detection and repair through machine learning

An automated system for detecting performance bugs in a program and for providing code recommendations to improve the performance of the program generates a code recommendation table from performance-related pull requests. The performance-related pull requests are identified in part from a classifier trained on semi-supervised data. A code recommendation table is generated from performance-related pull requests and is searched for similarly-improved code based on a set of difference features that includes structural and performance features of the before-code of a pull request that is not in the after-code.

Performance bug detection and code recommendation

Examples of the usage of a command of a command line interface includes the command with a set of parameters and corresponding parameter values. The examples are generated from telemetry data, which does not contain parameter values, and from web-based sources that may contain multiple parameter values. A machine learning model is used to predict the data type of a parameter value when the parameter is used with a particular command. The predicted data type is then used to select an appropriate parameter value for the example from multiple known parameter values or to generate a parameter value when no known parameter value exists.

Machine-generated examples of command-line commands with parameter values

Runtime errors in a source code program are detected in advance of execution by machine learning models. Features representing a context of a runtime error are extracted from source code programs to train a machine learning model, such as a random forest classifier, to predict the likelihood that a code snippet has a particular type of runtime error. The features are extracted from a syntax-type tree representation of each method in a program. A model is generated for distinct runtime errors, such as arithmetic overflow, and conditionally uninitialized variables.

Detection of runtime errors using machine learning

An automated command assistance tool is provided for a browser-enabled command line interface of a cloud service. The automated command assistance tool provides examples illustrating the correct syntax for commands used to manage the resources of a cloud service. The command assistance tool learns the syntax of a command from usage patterns found in telemetric data, scripts and user documentation and forms templates containing a command's usage pattern and related information. The templates are used to generate examples that respond to a user query for assistance with usage of a particular command.

Growing community of inventors

Kirkland, WA, United States of America

Roshanak Zilouchian Moghaddam