Packages

  • package root

    This is documentation for Mothra, a collection of Scala and Spark library functions for working with Internet-related data.

    This is documentation for Mothra, a collection of Scala and Spark library functions for working with Internet-related data. Some modules contain APIs of general use to Scala programmers. Some modules make those tools more useful on Spark data-processing systems.

    Please see the documentation for the individual packages for more details on their use.

    Scala Packages

    These packages are useful in Scala code without involving Spark:

    org.cert.netsa.data

    This package, which is collected as the netsa-data library, provides types for working with various kinds of information:

    org.cert.netsa.io.ipfix

    The netsa-io-ipfix library provides tools for reading and writing IETF IPFIX data from various connections and files.

    org.cert.netsa.io.silk

    To read and write CERT NetSA SiLK file formats and configuration files, use the netsa-io-silk library.

    org.cert.netsa.util

    The "junk drawer" of netsa-util so far provides only two features: First, a method for equipping Scala scala.collection.Iterators with exception handling. And second, a way to query the versions of NetSA libraries present in a JVM at runtime.

    Spark Packages

    These packages require the use of Apache Spark:

    org.cert.netsa.mothra.datasources

    Spark datasources for CERT file types. This package contains utility features which add methods to Apache Spark DataFrameReader objects, allowing IPFIX and SiLK flows to be opened using simple spark.read... calls.

    The mothra-datasources library contains both IPFIX and SiLK functionality, while mothra-datasources-ipfix and mothra-datasources-silk contain only what's needed for the named datasource.

    org.cert.netsa.mothra.analysis

    A grab-bag of analysis helper functions and example analyses.

    org.cert.netsa.mothra.functions

    This single Scala object provides Spark SQL functions for working with network data. It is the entirety of the mothra-functions library.

    Definition Classes
    root
  • package org
    Definition Classes
    root
  • package cert
    Definition Classes
    org
  • package netsa
    Definition Classes
    cert
  • package mothra
    Definition Classes
    netsa
  • package datasources

    This package contains the Mothra datasources, along with mechanisms for working with those datasources.

    This package contains the Mothra datasources, along with mechanisms for working with those datasources. The primary novel feature of these datasources is the fields mechanism.

    To use the IPFIX or SiLK data sources, you can use the following methods added by the implicit CERTDataFrameReader on DataFrameReader after importing from this package:

    import org.cert.netsa.mothra.datasources._
    val silkDF = spark.read.silkFlow()                                    // to read from the default SiLK repository
    val silkRepoDF = spark.read.silkFlow(repository="...")                // to read from an alternate SiLK repository
    val silkFilesDF = spark.read.silkFlow("/path/to/silk/files")          // to read from loose SiLK files
    val ipfixDF = spark.read.ipfix(repository="/path/to/mothra/data/dir") // for packed Mothra IPFIX data
    val ipfixS3DF = spark.read.ipfix(s3Repository="bucket-name")          // for packed Mothra IPFIX data from an S3 bucket
    val ipfixFilesDF = spark.read.ipfix("/path/to/ipfix/files")           // for loose IPFIX files

    (The additional methods are defined on the implicit class CERTDataFrameReader.)

    Using the fields method allows you to configure which SiLK or IPFIX fields you wish to retrieve. (This is particularly important for IPFIX data, as IPFIX files may contains many many possible fields organized in various ways.)

    import org.cert.netsa.mothra.datasources._
    val silkDF = spark.read.fields("sIP", "dIP").silkFlow(...)
    val ipfixDF = spark.read.fields("sourceIPAddress", "destinationIPAddress").ipfix(...)

    Both of these dataframes will contain only the source and destination IP addresses from the specified data sources. You may also provide column names different from the source field names:

    val silkDF = spark.read.fields("server" -> "sIP", "client" -> "dIP").silkFlow(...)
    val ipfixDF = spark.read.fields("server" -> "sourceIPAddress", "client" -> "destinationIPAddress").ipfix(...)

    You may also mix the mapped and the default names in one call:

    val df = spark.read.fields("sIP", "dIP", "s" -> "sensor").silkFlow(...)
    Definition Classes
    mothra
    See also

    IPFIX datasource

    SiLK flow datasource

  • package fields
    Definition Classes
    datasources
  • package ipfix

    A data source as defined by the Spark Data Source API for reading IPFIX records from Mothra data spools and from loose files.

    A data source as defined by the Spark Data Source API for reading IPFIX records from Mothra data spools and from loose files.

    You can use this by importing org.cert.netsa.mothra.datasources._ like this:

    import org.cert.netsa.mothra.datasources._
    val df1 = spark.read.ipfix("path/to/mothra/data/dir") // for packed Mothra IPFIX data
    val df2 = spark.read.ipfix("path/to/ipfix/files")     // for loose IPFIX files

    The IPFIX datasource uses the fields mechanism from org.cert.netsa.mothra.datasources. You can make use of this mechanism like these examples:

    import org.cert.netsa.mothra.datasources._
    val df1 = spark.read.fields(
      "startTime", "endTime", "sourceIPAddress", "destinationIPAddress"
    ).ipfix(...)
    
    val df2 = spark.read.fields(
      "startTime", "endTime", "TOS" -> "ipClassOfService"
    ).ipfix(...)

    with arbitrary sets of fields and field name mappings.

    Default Fields

    The default set of fields (defined in IPFIXFields.default) is:

    • "startTime" -> "func:startTime"
    • "endTime" -> "func:endTime"
    • "sourceIPAddress" -> "func:sourceIPAddress"
    • "sourcePort" -> "func:sourcePort"
    • "destinationIPAddress" -> "func:destinationIPAddress"
    • "destinationPort" -> "func:destinationPort"
    • "protocolIdentifier"
    • "observationDomainId"
    • "vlanId"
    • "reverseVlanId"
    • "silkAppLabel"
    • "packetCount" -> "packetTotalCount|packetDeltaCount"
    • "reversePacketCount" -> "reversePacketTotalCount|reversePacketDeltaCount"
    • "octetCount" -> "octetTotalCount|octetDeltaDcount"
    • "reverseOctetCount" -> "reverseOctetTotalCount|reverseOctetDeltaCount"
    • "initialTCPFlags"
    • "reverseInitialTCPFlags"
    • "unionTCPFlags"
    • "reverseUnionTCPFlags"

    Some of these defaults are defined simply as IPFIX Information Elements. For example, "protocolIdentifier" and "vlanId" are exactly the Information Elements that are named. No "right-hand-side" is given for these definitions, because the name of the field is the same as the name of the Information Element.

    Others have simple expressions. For example, packetCount is defined as "packetTotalCount|packetDeltaCount". This expressions means that the value should be found from the packetTotalCount IE, or if that is not set from the packetDeltaCount IE. This allows this field to be used regardless of which Information Element contains the data.

    Some others are derived in more complex ways from basic IPFIX fields. For example, the startTime field is produced using "func:startTime", which runs the "gauntlet of time" to determine the start time for a flow by whatever means possible. Other time fields are similarly defined.

    Some of the "func:..." fields are actually quite simple. For example, "func:sourceIPAddress", practically speaking, is the same as "sourceIPv4Address|sourceIPv6Address". However, these fields are defined using the func: extension mechanism so that partitioning on them is possible. (This restriction may be lifted in a future Mothra version.)

    Field Types

    The mappings between IPFIX types and Spark types are:

    • octetArray → Array[Byte]
    • unsigned8 → Short
    • unsigned16 → Int
    • unsigned32 → Long
    • unsigned64 → Long
    • signed8 → Byte
    • signed16 → Short
    • signed32 → Int
    • signed64 → Long
    • float32 → Float
    • float64 → Double
    • boolean → Boolean
    • macAddress → String
    • string → String
    • dateTimeSeconds → Timestamp
    • dateTimeMilliseconds → Timestamp
    • dateTimeMicroseconds → Timestamp
    • dateTimeNanoseconds → Timestamp
    • ipv4Address → String
    • ipv6Address → String

    IPFIX's basicList, subTemplateList, and subTemplateMultiList data types are handled differently.

    Field Expressions

    As noted above, field expressions may contain simple IPFIX Information Element names, or collections of names separated by pipe characters to indicate taking the first matching choice. This language has a number of other capabilities which are documented for now in the IPFIX field parser object.

    Functional Fields

    A number of pre-defined "functional fields" are available. Some of these combine other information elements in ways that the expression language cannot (applying the so-called "gauntlet of time", for example). Others provide support for the Mothra repository partitioning system. And finally, a few are for debugging purposes and provide high-level overviews of IPFIX records or point to file locations on disk.

    Function fields are all defined and described in the org.cert.netsa.mothra.datasources.ipfix.fields.func package.

    Definition Classes
    datasources
  • package silk
    Definition Classes
    datasources
  • package flow

    A data source as defined by the Spark Data Source API for reading SiLK records from SiLK data spools and from loose files.

    A data source as defined by the Spark Data Source API for reading SiLK records from SiLK data spools and from loose files.

    You can use this by importing org.cert.netsa.mothra.datasources._ like this:

    import org.cert.netsa.mothra.datasources._
    val df1 = spark.read.silkFlow() // to read from the default SiLK repository
    val df2 = spark.read.silkFlow("path/to/files") // to read from loose files
    val df3 = spark.read.silkFlow(..., repository="/path/to/repo") // to override the default repo
    val df4 = spark.read.silkFlow(..., configFile="/path/to/silk.conf") // to use a specific non-default silk.conf file

    The default SiLK data repository location is defined by the JAVA system property org.cert.netsa.mothra.datasources.silk.defaultRepository. The default configuration file is silk.conf under the default repository directory.

    If you don't have a SiLK data repository or a silk.conf file, you can still work with loose SiLK data files, however class, type, and sensor names will not be available. (Any numeric IDs in the input data will, however, still be usable.)

    The SiLK flow datasource uses the fields mechanism from org.cert.netsa.mothra.datasources. You can make use of this mechanism like these examples:

    import org.cert.netsa.mothra.datasources._
    spark.read.fields("sTime", "eTime", "sIP", "dIP").silkFlow(...)
    spark.read.fields("sTime", "sId" -> "sensorId").silkFlow(...)
    
    import org.cert.netsa.mothra.datasources.silk.flow.SilkFields
    spark.read.fields(SilkFields.default, "sId" -> "sensorId").silkFlow(...)

    with arbitrary sets of fields and field name mappings.

    See SilkFields for details about the default set of fields.

    SiLK field names match the names used by the SiLK rwcut tool when possible. Some fields which have variants in rwcut, such as duration and dur+msec, all mean the same thing in the SiLK flow datasource, to the maximum available resolution. Also note that unsigned numeric fields are generally one size larger in order to accommodate values too large to be represented in their base signed type. The full set of fields (along with aliases and Spark types) are listed below:

    • "application": Int
    • "attributes": Byte
    • "bytes": Long
    • "class": String
    • "dIP": String
    • "dPort": Int
    • "duration", "dur", "dur+msec": Long (in milliseconds)
    • "eTime", "eTime+msec": Timestamp
    • "filename": String
    • "flags": Byte
    • "flowType": String
    • "flowTypeId": Short
    • "iCode": Short
    • "iType": Short
    • "in": Int
    • "initialFlags": Byte
    • "isIPv6": Boolean
    • "memo": Short
    • "nhIP": String
    • "out": Int
    • "packets", "pkts": Long
    • "protocol": Short
    • "sIP": String
    • "sPort": Int
    • "sTime", "sTime+msec": Timestamp
    • "sensor": String
    • "sensorId": Int
    • "sessionFlags": Byte
    • "type": String
    • "filename": String

package silk

Package Members

  1. package flow

    A data source as defined by the Spark Data Source API for reading SiLK records from SiLK data spools and from loose files.

    A data source as defined by the Spark Data Source API for reading SiLK records from SiLK data spools and from loose files.

    You can use this by importing org.cert.netsa.mothra.datasources._ like this:

    import org.cert.netsa.mothra.datasources._
    val df1 = spark.read.silkFlow() // to read from the default SiLK repository
    val df2 = spark.read.silkFlow("path/to/files") // to read from loose files
    val df3 = spark.read.silkFlow(..., repository="/path/to/repo") // to override the default repo
    val df4 = spark.read.silkFlow(..., configFile="/path/to/silk.conf") // to use a specific non-default silk.conf file

    The default SiLK data repository location is defined by the JAVA system property org.cert.netsa.mothra.datasources.silk.defaultRepository. The default configuration file is silk.conf under the default repository directory.

    If you don't have a SiLK data repository or a silk.conf file, you can still work with loose SiLK data files, however class, type, and sensor names will not be available. (Any numeric IDs in the input data will, however, still be usable.)

    The SiLK flow datasource uses the fields mechanism from org.cert.netsa.mothra.datasources. You can make use of this mechanism like these examples:

    import org.cert.netsa.mothra.datasources._
    spark.read.fields("sTime", "eTime", "sIP", "dIP").silkFlow(...)
    spark.read.fields("sTime", "sId" -> "sensorId").silkFlow(...)
    
    import org.cert.netsa.mothra.datasources.silk.flow.SilkFields
    spark.read.fields(SilkFields.default, "sId" -> "sensorId").silkFlow(...)

    with arbitrary sets of fields and field name mappings.

    See SilkFields for details about the default set of fields.

    SiLK field names match the names used by the SiLK rwcut tool when possible. Some fields which have variants in rwcut, such as duration and dur+msec, all mean the same thing in the SiLK flow datasource, to the maximum available resolution. Also note that unsigned numeric fields are generally one size larger in order to accommodate values too large to be represented in their base signed type. The full set of fields (along with aliases and Spark types) are listed below:

    • "application": Int
    • "attributes": Byte
    • "bytes": Long
    • "class": String
    • "dIP": String
    • "dPort": Int
    • "duration", "dur", "dur+msec": Long (in milliseconds)
    • "eTime", "eTime+msec": Timestamp
    • "filename": String
    • "flags": Byte
    • "flowType": String
    • "flowTypeId": Short
    • "iCode": Short
    • "iType": Short
    • "in": Int
    • "initialFlags": Byte
    • "isIPv6": Boolean
    • "memo": Short
    • "nhIP": String
    • "out": Int
    • "packets", "pkts": Long
    • "protocol": Short
    • "sIP": String
    • "sPort": Int
    • "sTime", "sTime+msec": Timestamp
    • "sensor": String
    • "sensorId": Int
    • "sessionFlags": Byte
    • "type": String
    • "filename": String

Ungrouped