No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
dshell-defcon
web
.gitignore
Makefile
README.md
common.hh
indexer.cc
pcap2ap
split-flow.cc

README.md

PCAP Search

PCAP Search is a full-text substring search engine based on FM-index and other succinct data structures.

Installation

# dependencies of dshell
pip2 install --user dpkt pypcap

# dependencies of pcap2ap
# bc, inotify-tools

# dependencies of web/web.rb
gem install --user-install sinatra sinatra-contrib tilt sass slim coffee-script
# nodejs

Usage

mkdir -p /tmp/pcap/{all,eliza,wdub}

# create /tmp/pcap/all/a.cap with tcpdump/tshark
# create /tmp/pcap/wdub/a.cap with tcpdump/tshark

# Transform .cap files into .cap.ap files
# This intermediate format removes redundant metadata in PCAP/PCAPNG
# and is used for locating a specific packet in PCAP/PCAPNG files.
./pcap2ap -r /tmp/pcap &

# Transform .cap.ap files into .cap.ap.fm files
./indexer -r /tmp/pcap &

indexer search for .ap files in a directory, index them, and listen on a unix socket (/tmp/search.sock by default) to serve search queries.

Two types of search queries are provided: search and autocomplete.

Search

The simplest query is constructed with $'\0\0\0haystack' (zsh's quoting notation). indexer will print all the occurrences of haystack in all its indexed .ap files. You can restrict .ap files to be searched by providing the filename range: \0 $filename_begin \0 $filename_end \0 $query.

print is a builtin command in zsh.

query: \0 filename_begin \0 filename_end \0 query
result: filename \t offset \t context

# pattern is haystack
print -rn -- $'\0\0\0haystack' | socat -t 60 - unix:/tmp/search.sock

Autocomplete

A search query can be turned into an autocomplete query by supplying an offset number before the first \0.

query: offset \0 filename_begin \0 filename_end \0 query
result: filename \t offset \t context

# search, skip first 3 matches
print -rn -- $'3\0\0\0haystack' | socat -t 60 - unix:/tmp/search.sock

# search filenames F satisfying ("a" <= F <= "b"), skip first 5, pattern is "stack\0\0\1". \-escape is allowed
print -rn -- $'5\0a\0b\0ha\0stack\0\\0\\1' | socat -t 60 - unix:/tmp/search.sock

Web frontend

# change `PCAP_DIR = File.expand_path '/tmp/pcap'` in `web/web.rb`
web/web.rb

The web server will listen on 127.0.0.1:4568.

Internals

pcap2ap: extract TCP/UDP streams from .cap to .cap.ap

Implement a Dshell plugin dshell-defcon/dshell-decode to split a .cap to several streams and reassemble them into a .cap.ap file. A .cap.ap file is a concatenation of its streams, where each stream is composed of packets laid out in order. This format makes searching across packet boundary easier.

See ./dshell-defcon/README.md for detail.

pcap2ap is a shell wrapper of dshell-decode. It watches (inotify) .cap files in one or multiple directories and transforms them into .cap.ap files.

indexer: build a full-text index .cap.ap.fm for each .cap.ap and serve requests

indexer watches .fm indices in one or more directories and acts as a unix socket server supporing auto complete and search. For both types of queries, it scans watched .fm indices and locates the needle in the data files.

web: integrate indexer and the Dshell plugin

web/web.rb is a web application built upon Sinatra.

web/web.rb

.ap file specification

struct Ap {
  int32_t n_sessions;
  Session sessions[];
};

struct Session {
  int32_t n_packets;
  int32_t server_ip;
  int32_t server_port;
  int32_t client_ip;
  int32_t client_port;
  int32_t first_timestamp;
  int32_t last_timestamp;
  Packet packets[];
};

struct Packet {
  bool from_server;
  int32_t len;
};

.fm file specification

struct FM {
  char magic[8]; // GOODMEOW
  off_t len;
  // serialization of struct FMIndex
};