7. Zeek Integration¶

While Spicy itself remains application independent, transparent integration into Zeek has been a primary goal for its development. To facilitate adding new protocol and file analyzers to Zeek, there is a Zeek plugin that makes Spicy parsers accessible to Zeek’s processing pipeline. In the following, we dig deeper into how to use all of this.

Note

There are a number of pieces involved in creating a full Zeek analyzer, in particular if you want to distribute it as a Zeek package. To help you get started with that, Zeek’s package manager can create a skeleton Spicy package by running:

# zkg create --features=spicy-protocol-analyzer --packagedir <packagedir>

The generated files mark places that will need manual editing with TODO. See the tutorial for more on this.

7.1. Terminology¶

In Zeek, the term “analyzer” refers generally to a component that processes a particular protocol (“protocol analyzer”), file format (“file analyzer”), or low-level packet structure (“packet analyzer”). “Processing” here means more than just parsing content: An analyzer controls when it wants to be used (e.g., with connections on specific ports, or with files of a specific MIME type); what events to generate for Zeek’s scripting layer; and how to handle any errors occurring during parsing. While Spicy itself focuses just on the parsing part, the Spicy plugin makes it possible to provide the remaining pieces to Zeek, turning a Spicy parser into a full Zeek analyzer. That’s what we refer to as a “Spicy (protocol/file/packet) analyzer” for Zeek.

7.2. Installation¶

Since Zeek version 5.0, Spicy and Spicy plugin are by default bundled with Zeek. To confirm whether Spicy plugin is available you can inspect the list of installed plugin:

# zeek -N Zeek::Spicy
Zeek::Spicy - Support for Spicy parsers (*.hlto) (built-in)

If you see Zeek::Spicy listed, spicy-plugin is available, otherwise it needs to be installed manually. The recommended way to do so is through Zeek’s package manager zkg. If you have not yet installed zkg, follow its instructions.

You will need to have Spicy and Zeek installed as well of course. Before proceeding, make sure spicy-config and zeek-config are in your PATH:

# which spicy-config
/opt/spicy/bin/spicy-config

# which zeek-config
/usr/local/zeek/bin/zeek-config

7.2.1. Package Installation¶

The easiest way to install the plugin is through Zeek’s package manager:

# zkg install zeek/spicy-plugin

This will pull down the plugin’s package, compile and test the plugin, and then install and activate it. That process may take a bit to complete. To check afterwards that the plugin has become available, run zeek -N Zeek::Spicy, it should show output like this:

# zeek -N Zeek::Spicy
Zeek::Spicy - Support for Spicy parsers (*.hlto) (dynamic, version x.y.z)

By default, zkg will install the most recent release version of the plugin. If you want to install the current development version, use zkg install --version main zeek/spicy-plugin instead.

If you want to develop your own Spicy analyzers for Zeek, you will need a tool that comes with the plugin’s installation: spicyz. If you are using a recent version of zkg (>= 2.8.0), it’s easy to make the tool show up in your PATH: Either run zkg env (see this configuration advice) or update your PATH manually:

# export PATH=$(zkg config bin_dir):$PATH

If you are using an older version of zkg (including the version coming with Zeek 4.0), it’s a bit more difficult to find spicyz: it will be inside your zkg state directory at <state_dir>/clones/package/spicy-plugin/build/bin/spicyz. We recommend adding that directory to your PATH. (The state directory is usually either <zeek-prefix>/var/lib/zkg or ~/.zkg, depending on how you have set up zkg.)

7.2.2. Manual Installation¶

If you prefer, you can also compile the Zeek plugin yourself, outside of the package manager by cloning the plugin’s GitHub repository and building it through CMake. See the instructions in its README. This will install spicyz into <prefix>/bin.

Note

Developer’s note: It works to point ZEEK_PLUGIN_PATH directly to the plugin’s build directory, without installing it first. If you are building the plugin as part of the Spicy distribution, it will land in <build-directory>/zeek/spicy-plugin.

7.3. Interface Definitions (“evt files”)¶

Per above, a Spicy analyzer for Zeek does more than just parsing data. Accordingly, we need to tell the Zeek plugin a couple of additional pieces about analyzers we want it to provide to Zeek:

Analyzer setup: The plugin needs to know what type of analyzers we are creating, when we want Zeek to activate them, and what Spicy unit types to use as their parsing entry point.
Event definitions: We need to tell the Spicy plugin what Zeek events to provide and when to trigger them.

We define all of these through custom interface definition files that Spicy’s compiler for Zeek reads in. These files use an *.evt extension, and the following subsections discuss their content in more detail.

Generally, empty lines and comments starting with # are ignored in an *.evt.

Note

The syntax for *.evt files comes with some legacy pieces that aren’t particularly pretty. We may clean that up at some point.

7.3.1. Analyzer Setup¶

You can define protocol analyzers, packet analyzers and file analyzers in an *.evt file, per the following.

Protocol Analyzer

To define a protocol analyzer, add a new section to an *.evt file that looks like this:

protocol analyzer ANALYZER_NAME over TRANSPORT_PROTOCOL:
    PROPERTY_1,
    PROPERTY_2,
    ...
    PROPERTY_N;

Here, ANALYZER_NAME is a name to identify your analyzer inside Zeek. You can choose names arbitrarily as long as they are unique. As a convention, however, we recommend name with a spicy::* prefix (e.g., spicy::BitTorrent).

On the Zeek-side, through some normalization, these names automatically turn into tags added to Zeek’s Analyzer::Tag enum. For example, spicy::BitTorrent turns into Analyzer::ANALYZER_SPICY_BITTORRENT.

The analyzer’s name is also what goes into Zeek signatures to activate an analyzer DPD-style. If the name is spicy::BitTorrent, you’d write enable "spicy::BitTorrent" into the signature.

Note

Once you have made your analyzers available to Zeek (which we will discuss below), running zeek -NN Zeek::Spicy will show you a summary of what’s now available, including their Zeek-side names and tags.

TRANSPORT_PROTOCOL can be either tcp or udp, depending on the transport-layer protocol that your new analyzer wants to sit on top of.

Following that initial protocol analyzer ... line, a set of properties defines further specifics of your analyzer. The following properties are supported:

parse [originator|responder] with SPICY_UNIT
Specifies the top-level Spicy unit(s) the analyzer uses for parsing payload, with SPICY_UNIT being a fully-qualified Spicy-side type name (e.g. HTTP::Request). The unit type must have been declared as public in Spicy.

If originator is given, the unit is used only for parsing the connection’s originator-side payload; and if responder is given, only for responder-side payload. If neither is given, it’s used for both sides. In other words, you can use different units per side by specifying two properties parse originator with ... and parse responder with ....

port PORT or ports { PORT_1, ..., PORT_M }
Specifies one or more well-known ports for which you want Zeek to automatically activate your analyzer with corresponding connections. Each port must be specified in Spicy’s syntax for port constants (e.g., 80/tcp), or as a port range PORT_START-PORT_END where start and end port are port constants forming a closed interval. The ports’ transport protocol must match that of the analyzer.

Note

The plugin will also honor any %port meta data property that the responder-side SPICY_UNIT may define (as long as the attribute’s direction is not originator).

replaces ANALYZER_NAME
Disables an existing analyzer that Zeek already provides internally, allowing you to replace a built-in analyzer with a new Spicy version. ANALYZER_NAME is the Zeek-side name of the analyzer. To find that name, inspect the output of zeek -NN for available analyzers:
# zeek -NN | grep '\[Analyzer\]'
...
[Analyzer] SMTP (ANALYZER_SMTP, enabled)
...
Here, SMTP is the name you would write into replaces to disable the built-in SMTP analyzer.

As a full example, here’s what a new HTTP analyzer could look like:

protocol analyzer spicy::HTTP over TCP:
    parse originator with HTTP::Requests,
    parse responder with HTTP::Replies,
    port 80/tcp,
    replaces HTTP;

Packet Analyzer

Defining packet analyzers works quite similar to protocol analyzers through *.evt sections like this:

packet analyzer ANALYZER_NAME:
    PROPERTY_1,
    PROPERTY_2,
    ...
    PROPERTY_N;

Here, ANALYZER_NAME is again a name to identify your analyzer inside Zeek. On the Zeek-side, the name will be added to Zeek’s PacketAnalyzer::Tag enum.

Packet analyzers support the following properties:

parse with SPICY_UNIT
Specifies the top-level Spicy unit the analyzer uses for parsing each packet, with SPICY_UNIT being a fully-qualified Spicy-side type name. The unit type must have been declared as public in Spicy.

replaces ANALYZER_NAME
Disables an existing packet analyzer that Zeek already provides internally, allowing you to replace a built-in analyzer with a new Spicy version. ANALYZER_NAME is the Zeek-side name of the existing analyzer. To find that name, inspect the output of zeek -NN for available analyzers:
# zeek -NN | grep '\[Packet Analyzer\]'
...
[Packet Analyzer] Ethernet (ANALYZER_ETHERNET)
...
Here, Ethernet is the name you would give to replaces to disable the built-in Ethernet analyzer.

When replacing an existing packet analyzer, you still need to configure your new analyzer on the Zeek side through a zeek_init() event handler. See below for more.

Note

This feature requires Zeek >= 5.2.

As a full example, here’s what a new packet analyzer could look like:

packet analyzer spicy::RawLayer:
    parse with RawLayer::Packet;

In addition to the Spicy-side configuration, packet analyzers also need to be registered with Zeek inside a zeek_init event handler; see the Zeek documentation for more. The -NN output shows your new analyzer’s ANALYZER_* tag to use.

Note

Depending on relative loading order between your analyzer and Zeek’s scripts, in some situations you may not have access to the ANALYZER_* tag referring to your new analyzer. If you run into this, use PacketAnalyzer::try_register_packet_analyzer_by_name to register your Spicy analyzer by name (as shown by -NN). Example:

event zeek_init()
    {
    if ( ! PacketAnalyzer::try_register_packet_analyzer_by_name("Ethernet", 0x88b5, "spicy::RawLayer") )
            Reporter::error("unknown analyzer name used");
    }

File Analyzer

Defining file analyzers works quite similar to protocol analyzers, through *.evt sections like this:

file analyzer ANALYZER_NAME:
    PROPERTY_1,
    PROPERTY_2,
    ...
    PROPERTY_N;

Here, ANALYZER_NAME is again a name to identify your analyzer inside Zeek. On the Zeek-side, the name will be added to Zeek’s Files::Tag enum.

File analyzers support the following properties:

parse with SPICY_UNIT
Specifies the top-level Spicy unit the analyzer uses for parsing file content, with SPICY_UNIT being a fully-qualified Spicy-side type name. The unit type must have been declared as public in Spicy.

mime-type MIME-TYPE
Specifies a MIME type for which you want Zeek to automatically activate your analyzer when it sees a corresponding file on the network. The type is specified in standard type/subtype notion, without quotes (e.g., image/gif).

Note

The plugin will also honor any %mime-type meta data property that the SPICY_UNIT may define.

Note

Keep in mind that Zeek identifies MIME types through “content sniffing” (i.e., similar to libmagic), and usually not by protocol-level headers (e.g., not through HTTP’s Content-Type header). If in doubt, examine files.log for what it records as a file’s type.

replaces ANALYZER_NAME
Disables an existing file analyzer that Zeek already provides internally, allowing you to replace a built-in analyzer with a new Spicy version. ANALYZER_NAME is the Zeek-side name of the analyzer. To find that name, inspect the output of zeek -NN for available analyzers:
# zeek -NN | grep '\[File Analyzer\]'
...
[File Analyzer] PE (ANALYZER_PE, enabled)
...
Here, PE is the name you would write into replaces to disable the built-in PE analyzer.

Note

This feature requires Zeek >= 4.1.

As a full example, here’s what a new GIF analyzer could look like:

file analyzer spicy::GIF:
    parse with GIF::Image,
    mime-type image/gif;

7.3.2. Event Definitions¶

To define a Zeek event that you want the Spicy plugin to trigger, you add lines of the form:

on HOOK_ID -> event EVENT_NAME(ARG_1, ..., ARG_N);

on HOOK_ID if COND -> event EVENT_NAME(ARG_1, ..., ARG_N);

The Zeek plugin automatically derives from this everything it needs to register new events with Zeek, including a mapping of the arguments’ Spicy types to corresponding Zeek types. More specifically, these are the pieces going into such an event definition:

on HOOK_ID

A Spicy-side ID that defines when you want to trigger the event. This works just like an on ... unit hook, and you can indeed use anything here that Spicy supports for those as well (except container hooks). So, e.g., on HTTP::Request::%done triggers an event whenever a HTTP::Request unit has been fully parsed, and on HTTP::Request::uri leads to an event each time the uri field has been parsed. (In the former example you may skip the %done, actually: on HTTP::Request implicitly adds it.)

EVENT_NAME

The Zeek-side name of the event you want to generate, preferably including a namespace (e.g., http::request).

ARG_1, ..., ARG_N

Arguments to pass to the event, given as arbitrary Spicy expressions. Each expression will be evaluated within the context of the unit that the on ... triggers on, similar to code running inside the body of a corresponding unit hook. That means the expression has access to self for accessing the unit instance that’s currently being parsed.

The Spicy type of the expression determines the Zeek-side type of the corresponding event argument. Most Spicy types translate over pretty naturally, the following summarizes the translation:

Type Conversion from Spicy to Zeek¶
Spicy Type	Zeek Type	Notes
`addr`	`addr`
`bool`	`bool`
`enum { ... }`	`enum { ... }`	[1]
`int(8\|16\|32\|64)`	`int`
`interval`	`interval`
`list<T>`	`vector of T`
`map<V,K>`	`table[V] of K`
`optional<T>`	`T`	[2]
`port`	`port`
`real`	`double`
`set<T>`	`set[T]`
`string`	`string`
`time`	`time`
`tuple<T_1, ... ,T_N>`	`record { T1, ..., T_N }`	[3]
`uint(8\|16\|32\|64)`	`count`
`vector<T>`	`vector of T`

Note

[1]

A corresponding Zeek-side enum type is automatically created. See below for more.

[2]

The optional value must have a value, otherwise a runtime exception will be thrown.

[3]

Must be mapped to a Zeek-side record type with matching fields.

If a tuple element is mapped to a record field with a &default or &optional attribute, a couple special cases are supported:

If the expression evaluates to Null, the record field is left unset.

If the element’s expression uses the .? operator and that fails to produce a value, the record field is likewise left unset.

In addition to full Spicy expressions, there are three reserved IDs with specific meanings when used as arguments:

$conn
Refers to the connection that’s currently being processed by Zeek. On the Zeek-side this will turn into a parameter of Zeek type connection. This ID can be used only with protocol analyzers.

$file
Refers to the file that’s currently being processed by Zeek. On the Zeek-side this will turn into a parameter of Zeek type fa_file. This ID can be used only with file analyzers.

$packet
Refers to the packet that’s currently being processed by Zeek. On the Zeek-side this will turn into a parameter of Zeek type raw_pkt_hdr, with any fields filled in that have already been parsed by Zeek’s built-in analyzers. This ID can be used only with packet analyzers. (Note that instantiation of raw_pkt_hdr values can be relatively expensive on the Zeek side, so best to limit usage of this ID to a small part of the overall traffic.)

$is_orig
A boolean indicating if the data currently being processed is coming from the originator (True) or responder (False) of the underlying connection. This turns into a corresponding boolean value on the Zeek side. This ID can be used only with protocol analyzers.

Note

Some tips:

If you want to force a specific type on the Zeek-side, you have a couple of options:
1. Spicy may provide a cast operator from the actual type into the desired type (e.g., cast<uint64>(..)).
2. Argument expressions have access to global functions defined in the Spicy source files, so you can write a conversion function taking an argument with its original type and returning it with the desired type.
List comprehension can be convenient to fill Zeek vectors: [some_func(i) for i in self.my_list].

if COND

If given, events are only generated if the expression COND evaluates to true. Just like event arguments, the expression is evaluated in the context of the current unit instance and has access to self.

7.3.3. Exporting Types¶

As we discuss above, the type of each event argument maps over to a corresponding Zeek type. On the Zeek side, that corresponding type needs to be known by Zeek. That is always the case for built-in atomic types (per the table:ref:the conversion table <zeek-event-arg-types>), but it can turn out more challenging to achieve for custom types, such as enum and record, for which you would normally need to create matching type declarations in your Zeek scripts. While that’s not necessarily hard, but it can become quite cumbersome.

Fortunately, the Spicy plugin can help: for most types, it can instantiate corresponding Zeek types automatically as it loads the corresponding analyzer. While you will never actually see the Zeek-side type declarations, they will be available inside your Zeek scripts as if you had typed them out yourself–similar to other types that are built into Zeek itself.

To have the plugin create a Zeek type for your analyzer automatically, you need to export the Spicy type in your EVT file. The syntax for that is:

export SPICY_ID [as ZEEK_ID];

Here, SPICY_ID is the fully-scoped type ID on the Spicy side, and ZEEK_ID is the fully-scoped type ID you want in Zeek. If you leave out the as ... part, the Zeek name will be the same as the Spicy name, including its namespace. For example, say you have a Spicy unit TFTP::Request. Adding export TFTP::Request; to your EVT file will make a record type of the same name, and with the same fields, available in Zeek. If you instead use export TFTP::Request as TheOtherTFTP::Request, it will be placed into a different namespace instead.

Exporting types generally works for most Spicy types as long as there’s an ID associated with them in your Spicy code. However, exporting is most helpful with user-defined types, such as enum and unit, because it can save you quite a bit of typing there. We discuss the more common type conversions in more detail below.

To confirm the types made available to Zeek, you can see all exports in the output of zeek -NN. With our 2nd TFTP::Request example, that looks like this:

[...]
# zeek -NN Zeek::Spicy
Zeek::Spicy - Support for Spicy parsers (*.hlto)
    [Type] TheOtherTFTP::Request
[...]

Note

Most, but not all, types can be exported automatically. For example, self-recursive types are currently not supported. Generally, if you run into trouble exporting a type, you can always fall back to declaring a corresponding Zeek version yourself in your Zeek script. Consider the export mechanism as a convenience feature that helps avoid writing a lot of boiler plate code in common cases.

7.3.3.1. Enum Types¶

When you export a Spicy enum type, the Spicy plugin creates a corresponding Zeek enum type. For example, assume the following Spicy declaration:

module Test;

type MyEnum = enum {
    A = 83,
    B = 84,
    C = 85
};

Using export Test::MyEnum;, the plugin will create the equivalent of the following Zeek type for use in your scripts:

module Test;

export {

  type MyEnum: enum {
      MyEnum_A = 83,
      MyEnum_B = 84,
      MyEnum_A = 85,
      MyEnum_Undef = -1
  };

}

(The odd naming is due to ID limitations on the Zeek side.)

Note

For backward compatibility, the enum type comes with an additional property: all public enum types are automatically exported, even without adding any export to your EVT file. This feature may go away at some point, and we suggest to not rely on it on new code; always use export for enum types as well.

7.3.3.2. Unit Types¶

When you export a Spicy unit type, the Spicy plugin creates a corresponding Zeek record type. For example, assume the following Spicy declaration:

module Test;

type MyRecord = unit {
    a: bytes &size=4;
    b: uint16;

    var c: bool;
};

Using export Test::MyRecord;, the plugin will then create the equivalent of the following Zeek type for use in your scripts:

module Test;

export {

  type MyRecord: record {
    a: string &optional;
    b: count &optional;
    c: bool;
  };

}

The individual fields map over just like event arguments do, following the the table above. For aggregate types, this works recursively: if, e.g., a field is itself of unit type and that type has been exported as well, the plugin will map it over accordingly to the corresponding record type. Note that such dependent types must be exported first in the EVT file for this to work. As a result, you cannot export self-recursive unit types.

As you can see in the example, unit fields are always declared as optional on the Zeek-side, as they may not have been set during parsing. Unit variables are non-optional by default, unless declared as &optional in Spicy.

7.3.3.3. Struct Types¶

A Spicy struct type maps over to Zeek in the same way as unit types do, treating each field like a unit variable. See there for more information.

7.3.3.4. Tuple Types¶

A Spicy tuple type maps over to a Zeek record similar to how unit types do, treating each tuple element like a unit variable. See there for more information.

Exporting works only for tuple types that declare names for all their elements.

7.3.4. Importing Spicy Modules¶

Code in an *.evt file may need access to additional Spicy modules, such as when expressions for event parameters call Spicy functions defined elsewhere. To make a Spicy module available, you can insert import statements into the *.evt file that work just like in Spicy code:

import NAME
Imports Spicy module NAME.

import NAME from X.Y.Z;
Searches for the module NAME (i.e., for the filename NAME.spicy) inside a sub-directory X/Y/Z along the search path, and then imports it.

7.3.5. Conditional Compilation¶

*.evt files offer the same basic form of conditional compilation through @if/@else/@endif blocks as Spicy scripts. The Zeek plugin makes two additional identifiers available for testing to both *.evt and *.spicy code:

HAVE_ZEEK
Always set to 1 by the plugin. This can be used for feature testing from Spicy code to check if it’s being compiled for Zeek.

ZEEK_VERSION
The numerical Zeek version that’s being compiled for (see zeek -e 'print Version::number').

This is an example bracketing code by Zeek version in an EVT file:

@if ZEEK_VERSION < 30200
    <EVT code for Zeek versions older than 3.2>
@else
    <EVT code for Zeek version 3.2 or newer>
@endif

7.4. Compiling Analyzers¶

Once you have the *.spicy and *.evt source files for your new analyzer, you need to precompile them into an *.hlto object file containing their final executable code, which Zeek will then use. To do that, pass the relevant *.spicy and *.evt files to spicyz, then have Zeek load the output. To repeat the example from the Getting Started guide:

# spicyz -o my-http-analyzer.hlto my-http.spicy my-http.evt
# zeek -Cr request-line.pcap my-http-analyzer.hlto my-http.zeek
Zeek saw from 127.0.0.1: GET /index.html 1.0

Instead of providing the precompiled analyzer on the Zeek command line, you can also copy them into ${prefix}/lib/spicy/Zeek_Spicy/modules. The Spicy plugin will automatically load any *.hlto object files it finds there. In addition, the plugin also scans Zeek’s plugin directory for *.hlto files. Alternatively, you can override both of those locations by setting the environment variable ZEEK_SPICY_MODULE_PATH to a set of colon-separated directories to search instead. The plugin will then only look there. In all cases, the plugin searches any directories recursively, so it will find *.hlto also if they are nested in subfolders.

Run spicyz -h to see some additional options it provides, which are similar to spicy-driver.

7.5. Controlling Zeek from Spicy¶

Spicy grammars can import a provided library module zeek to gain access to Zeek-specific functions that call back into Zeek’s processing:

function zeek::confirm_protocol()

[Deprecated] Triggers a DPD protocol confirmation for the current connection.

This function has been deprecated and will be removed. Use spicy::accept_input instead, which will have the same effect with Zeek.

function zeek::reject_protocol(reason: string)

[Deprecated] Triggers a DPD protocol violation for the current connection.

This function has been deprecated and will be removed. Use spicy::decline_input instead, which will have the same effect with Zeek.

function zeek::weird(id: string, addl: string = "") : &cxxname="spicy::zeek::rt::weird";

Reports a “weird” to Zeek. This should be used with similar semantics as in Zeek: something quite unexpected happening at the protocol level, which however does not prevent us from continuing to process the connection.

id: the name of the weird, which (just like in Zeek) should be a static string identifying the situation reported (e.g., unexpected_command).

addl: additional information to record along with the weird

function zeek::is_orig() : bool

Returns true if we’re currently parsing the originator side of a connection.

function zeek::uid() : string

Returns the current connection’s UID.

function zeek::conn_id() : tuple<orig_h: addr, orig_p: port, resp_h: addr, resp_p: port>

Returns the current connection’s 4-tuple ID.

function zeek::flip_roles()

Instructs Zeek to flip the directionality of the current connection.

function zeek::number_packets() : uint64

Returns the number of packets seen so far on the current side of the current connection.

function zeek::protocol_begin(analyzer: optional<string> = Null)

Adds a Zeek-side child protocol analyzer to the current connection.

analyzer: type of analyzer to instantiate, specified through its Zeek-side name (similar to what Zeek’s signature action enable takes); if not specified, Zeek will perform its usual dynamic protocol detection to figure out how to parse the data (the latter will work only for TCP protocols, though.)

function zeek::protocol_data_in(is_orig: bool, data: bytes)

Forwards protocol data to all previously instantiated Zeek-side child protocol analyzers.

is_orig: true to feed the data to the child’s originator side, false for the responder data: chunk of data to forward to child analyzer

Note that due to Zeek limitations, any data passed through this function will always be forwarded to all currently active child analyzers.

function zeek::protocol_gap(is_orig: bool, offset: uint64, len: uint64)

Signals a gap in input data to all previously instantiated Zeek-side child protocol analyzers.

is_orig: true to signal gap to the child’s originator side, false for the responder offset: start offset of gap in input stream len: size of gap

Note that due to Zeek limitations, any gaps signaled through this function will always be forwarded to all currently active child analyzers.

function zeek::protocol_end()

Signals end-of-data to all previously instantiated Zeek-side child protocol analyzers and removes them.

function zeek::file_begin(mime_type: optional<string> = Null) : string

Signals the beginning of a file to Zeek’s file analysis, associating it with the current connection. Optionally, a mime type can be provided. It will be passed on to Zeek’s file analysis framework. Returns the Zeek-side file ID of the new file.

function zeek::fuid() : string

Returns the current file’s FUID.

function zeek::terminate_session()

Terminates the currently active Zeek-side session, flushing all state. Any subsequent activity will start a new session from scratch. This can only be called from inside a protocol analyzer.

function zeek::file_set_size(size: uint64, fid: optional<string> = Null)

Signals the expected size of a file to Zeek’s file analysis.

size: expected size of file fid: Zeek-side ID of the file to operate on; if not given, the file started by the most recent file_begin() will be used

function zeek::file_data_in(data: bytes, fid: optional<string> = Null)

Passes file content on to Zeek’s file analysis.

data: chunk of raw data to pass into analysis fid: Zeek-side ID of the file to operate on; if not given, the file started by the most recent file_begin() will be used

function zeek::file_data_in_at_offset(data: bytes, offset: uint64, fid: optional<string> = Null)

Passes file content at a specific offset on to Zeek’s file analysis.

data: chunk of raw data to pass into analysis offset: position in file where data starts fid: Zeek-side ID of the file to operate on; if not given, the file started by the most recent file_begin() will be used

function zeek::file_gap(offset: uint64, len: uint64, fid: optional<string> = Null)

Signals a gap in a file to Zeek’s file analysis.

offset: position in file where gap starts len: size of gap fid: Zeek-side ID of the file to operate on; if not given, the file started by the most recent file_begin() will be used

function zeek::file_end(fid: optional<string> = Null)

Signals the end of a file to Zeek’s file analysis.

fid: Zeek-side ID of the file to operate on; if not given, the file started by the most recent file_begin() will be used

function zeek::forward_packet(identifier: uint32)

Inside a packet analyzer, forwards what data remains after parsing the top-level unit on to another analyzer. The index specifies the target, per the current dispatcher table.

function zeek::network_time() : time

Gets the network time from Zeek.

7.6. Dynamic Protocol Detection (DPD)¶

Spicy protocol analyzers support Zeek’s Dynamic Protocol Detection (DPD), i.e., analysis independent of any well-known ports. To use that with your analyzer, add two pieces:

A Zeek signature to activate your analyzer based on payload patterns. Just like with any of Zeek’s standard analyzers, a signature can activate a Spicy analyzer through the enable "<name>" keyword. The name of the analyzer comes out of the EVT file: it is the ANALYZER_NAME with the double colons replaced with an underscore (e.g., spicy::HTTP turns into enable "spicy_HTTP".
You should call spicy::accept_input() from a hook inside your grammar at a point when the parser can be reasonably certain that it is processing the expected protocol. Optionally, you may also call spicy::decline_input() when you’re sure the parser is not parsing the right protocol. However, the Zeek plugin will also trigger this automatically whenever your parser aborts with an error.

7.7. Configuration¶

7.7.1. Options¶

The Spicy plugin provides a set of script-level options to tune its behavior. These all live in the Spicy:: namespace:

    # Constant for testing if Spicy is available.
    const available = T;

    ## Show output of Spicy print statements.
    const enable_print = F &redef;

    ## abort() instead of throwing HILTI exceptions.
    const abort_on_exceptions = F &redef;

    ## Include backtraces when reporting unhandled exceptions.
    const show_backtraces = F &redef;

    ## Maximum depth of recursive file analysis (Spicy analyzers only)
    const max_file_depth: count = 5 &redef;

7.7.2. Functions¶

The Spicy plugin also adds the following new built-in functions to Zeek, which likewise live in the Spicy:: namespace:

    ## Enable a specific Spicy protocol analyzer if not already active. If this
    ## analyzer replaces an standard analyzer, that one will automatically be
    ## disabled.
    ##
    ## tag: analyzer to toggle
    ##
    ## Returns: true if the operation succeeded
    global enable_protocol_analyzer: function(tag: Analyzer::Tag) : bool;

    ## Disable a specific Spicy protocol analyzer if not already inactive. If
    ## this analyzer replaces an standard analyzer, that one will automatically
    ## be re-enabled.
    ##
    ## tag: analyzer to toggle
    ##
    ## Returns: true if the operation succeeded
    global disable_protocol_analyzer: function(tag: Analyzer::Tag) : bool;


    ## Enable a specific Spicy file analyzer if not already active. If this
    ## analyzer replaces an standard analyzer, that one will automatically be
    ## disabled.
    ##
    ## tag: analyzer to toggle
    ##
    ## Returns: true if the operation succeeded
    global enable_file_analyzer: function(tag: Files::Tag) : bool;

    ## Disable a specific Spicy file analyzer if not already inactive. If
    ## this analyzer replaces an standard analyzer, that one will automatically
    ## be re-enabled.
    ##
    ## tag: analyzer to toggle
    ##
    ## Returns: true if the operation succeeded
    global disable_file_analyzer: function(tag: Files::Tag) : bool;

7.8. Debugging¶

If Zeek doesn’t seem to be doing the right thing with your Spicy analyzer, there are several ways to debug what’s going on. To facilitate that, compile your analyzer with spicyz -d and, if possible, use a debug version of Zeek (i.e., build Zeek with ./configure --enable-debug).

If your analyzer doesn’t seem to be active at all, first make sure Zeek actually knows about it: It should show up in the output of zeek -NN Zeek::Spicy. If it doesn’t, you might not have been using the right *.spicy or *.evt files when precompiling, or Zeek is not loading the *.hlto file. Also check your *.evt if it defines your analyzer correctly.

If Zeek knows about your analyzer and just doesn’t seem to activate it, double-check that ports or MIME types are correct in the *.evt file. If you’re using a signature instead, try a port/MIME type first, just to make sure it’s not a matter of signature mismatches.

If there’s nothing obviously wrong with your source files, you can trace what the plugin is compiling by running spicyz with -D zeek. For example, reusing the HTTP example from the Getting Started guide:

# spicyz -D zeek my-http.spicy my-http.evt -o my-http.hlt
[debug/zeek] Loading Spicy file "/Users/robin/work/spicy/main/tests/spicy/doc/my-http.spicy"
[debug/zeek] Loading EVT file "/Users/robin/work/spicy/main/doc/examples/my-http.evt"
[debug/zeek] Loading events from /Users/robin/work/spicy/main/doc/examples/my-http.evt
[debug/zeek]   Got protocol analyzer definition for spicy_MyHTTP
[debug/zeek]   Got event definition for MyHTTP::request_line
[debug/zeek] Running Spicy driver
[debug/zeek]   Got unit type 'MyHTTP::Version'
[debug/zeek]   Got unit type 'MyHTTP::RequestLine'
[debug/zeek] Adding protocol analyzer 'spicy_MyHTTP'
[debug/zeek] Adding Spicy hook 'MyHTTP::RequestLine::0x25_done' for event MyHTTP::request_line
[debug/zeek] Done with Spicy driver

You can see the main pieces in there: The files being loaded, unit types provided by them, analyzers and events being created.

If that all looks as expected, it’s time to turn to the Zeek side and see what it’s doing at runtime. You’ll need a debug version of Zeek for that, as well as a small trace with traffic that you expect your analyzer to process. Run Zeek with -B dpd (or -B file_analysis if you’re debugging a file analyzer) on your trace to record the analyzer activity into debug.log. For example, with the same HTTP example, we get:

# zeek -B dpd -Cr request-line.pcap my-http.hlto
# cat debug.log
[dpd] Registering analyzer SPICY_MYHTTP for port 12345/1
[...[
[dpd] Available analyzers after zeek_init():
[...]
[dpd]     spicy_MyHTTP (enabled)
[...]
[dpd] Analyzers by port:
[dpd]     12345/tcp: SPICY_MYHTTP
[...]
[dpd] TCP[5] added child SPICY_MYHTTP[7]
[dpd] 127.0.0.1:59619 > 127.0.0.1:12345 activated SPICY_MYHTTP analyzer due to port 12345
[...]
[dpd] SPICY_MYHTTP[7] DeliverStream(25, T) [GET /index.html HTTP/1.0\x0a]
[dpd] SPICY_MYHTTP[7] EndOfData(T)
[dpd] SPICY_MYHTTP[7] EndOfData(F)

The first few lines show that Zeek’s analyzer system registers the analyzer as expected. The subsequent lines show that the analyzer gets activated for processing the connection in the trace, and that it then receives the data that we know indeed constitutes its payload, before it eventually gets shutdown.

To see this from the plugin’s side, set the zeek debug stream through the HILTI_DEBUG environment variable:

# HILTI_DEBUG=zeek zeek -Cr request-line.pcap my-http.hlto
[zeek] Have Spicy protocol analyzer spicy_MyHTTP
[zeek] Registering Protocol::TCP protocol analyzer spicy_MyHTTP with Zeek
[zeek]   Scheduling analyzer for port 12345/tcp
[zeek] Done with post-script initialization
[zeek] [SPICY_MYHTTP/7/orig] initial chunk: |GET /index.html HTTP/1.0\\x0a| (eod=false)
[zeek] [SPICY_MYHTTP/7/orig] -> event MyHTTP::request_line($conn, GET, /index.html, 1.0)
[zeek] [SPICY_MYHTTP/7/orig] done with parsing
[zeek] [SPICY_MYHTTP/7/orig] parsing finished, skipping further originator payload
[zeek] [SPICY_MYHTTP/7/resp] no unit specified for parsing
[zeek] [SPICY_MYHTTP/7/orig] skipping end-of-data delivery
[zeek] [SPICY_MYHTTP/7/resp] no unit specified for parsing
[zeek] [SPICY_MYHTTP/7/orig] skipping end-of-data delivery
[zeek] [SPICY_MYHTTP/7/resp] no unit specified for parsing

After the initial initialization, you see the data arriving and the event being generated for Zeek. The plugin also reports that we didn’t define a unit for the responder side—which we know in this case, but if that appears unexpectedly you probably found a problem.

So we know now that our analyzer is receiving the anticipated data to parse. At this point, we can switch to debugging the Spicy side through the usual mechanisms. In particular, setting HILTI_DEBUG=spicy tends to be helpful:

# HILTI_DEBUG=spicy zeek -Cr request-line.pcap my-http.hlto
[spicy] MyHTTP::RequestLine
[spicy]   method = GET
[spicy]   anon_2 =
[spicy]   uri = /index.html
[spicy]   anon_3 =
[spicy]   MyHTTP::Version
[spicy]     anon = HTTP/
[spicy]     number = 1.0
[spicy]   version = [$number=b"1.0"]
[spicy]   anon_4 = \n

If everything looks right with the parsing, and the right events are generated too, then the final part is to check out the events that arrive on the Zeek side. To get Zeek to see an event that the plugin raises, you need to have at least one handler implemented for it in one of your Zeek scripts. You can then load Zeek’s misc/dump-events to see them as they are being received, including their full Zeek-side values:

# zeek -Cr request-line.pcap my-http.hlto misc/dump-events
[...]
1580991211.780489 MyHTTP::request_line
              [0] c: connection      = [id=[orig_h=127.0.0.1, orig_p=59619/tcp, ...] ...]
              [1] method: string     = GET
              [2] uri: string        = /index.html
              [3] version: string    = 1.0
[...]