Browse Prior Art Database

PROACTIVE BUG ALERTS

IP.com Disclosure Number: IPCOM000248535D
Publication Date: 2016-Dec-14
Document File: 7 page(s) / 557K

Publishing Venue

The IP.com Prior Art Database

Related People

Amod Augustin: AUTHOR [+4]

Abstract

Systems and methods described herein provide proactive bug vulnerability alerts to customers based on the customer environment to enable customers to take precautionary measures to mitigate bug vulnerability. This system provides an end-to-end solution useful to improve customer satisfaction.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 39% of the total text.

Copyright 2016 Cisco Systems, Inc. 1

PROACTIVE BUG ALERTS

AUTHORS: Amod Augustin

Gyana Dash Jinkle Jose

Milind Naphade

CISCO SYSTEMS, INC.

ABSTRACT

Systems and methods described herein provide proactive bug vulnerability alerts

to customers based on the customer environment to enable customers to take

precautionary measures to mitigate bug vulnerability. This system provides an end-to-end

solution useful to improve customer satisfaction.

DETAILED DESCRIPTION

During software development, bugs often arise when situations causing particular

corner cases are not tested. Because it may not be possible to test every scenario, bugs are

frequently discovered by customers or internal testing after an Operating System (OS)

version release. Individual customer issues relating to bugs may be resolved by a

customer assistance program (CAP), a software maintenance update (SMU), or in a new

version.

It is desirable to propagate knowledge regarding bug vulnerability to all

customers, whether they use a CAP or not. Techniques described herein quickly identify

the cause of the bug, trace the configuration (config) commands causing the bug to

trigger, and communicate appropriate actions to the customers. These techniques map to

commands rather than to features. Once the cause of the bug is recorded, this knowledge

is applied to all customers via a portal or any contract agreement.

Commands, which that are enabled in a customer device configuration file, are

learned using natural language processing (NLP) technology. The commands are mapped

to bugs by correlating data across multiple sources (e.g., bugs, product documentation,

Copyright 2016 Cisco Systems, Inc. 2

etc.) using machine learning techniques. These techniques provide proactive bug alerts to

customers.

Figure 1 below illustrates an overall flow of a process for providing a proactive

bug alert.

Figure 1

One approach uses NLP to extract the information from the bug. The following is

an example step-by-step process.

Step 1: Assuming access to relevant information relating to the bug is available, the

relevant information, for example title, description, release notes, bug description (if

available), all possible attachments, and metadata (e.g., status, project, component, etc.),

is extracted. Basic NLP is performed to clean up the data, including stemming,

lemmatization, n-gram, and Word2vec. In an example, only bugs relevant to the metadata

of the customers are extracted.

Step 2: The command is extracted from the bug data. The model may be trained with the

relevant ontology, command synonym, and configs. A command synonym may be

created via Word2vec from the bug data, or product manual, thereby providing similar

terms (for example, SNMP = [mib, oid, SNMP, ...]).

Copyright 2016 Cisco Systems, Inc. 3

Step 3: The topics are prioritized. Once the topics are extracted from the bugs, they are

prioritized based on the weight, where the keyword is present in the bug data. In an

example, the more frequently a key...