Skip to content

Files

Latest commit

peace-makerpeace-maker
Update TODO
Nov 16, 2023
fb9999e · Nov 16, 2023

History

History
90 lines (86 loc) · 5.95 KB

TODO.md

File metadata and controls

90 lines (86 loc) · 5.95 KB

pkappa2 TODO

server:

  • support http
  • support http/1 websocket, including compression
  • support http/2
  • support http/2 websocket
  • support quic
  • support ocsp
  • support SignalR
  • support is:started|finished
  • support pcap groups, they have their own indexes & snapshots and may only be combined with packets in the same group
  • fix ip4 defragmentation (snapshottable, list of packets that are source for a reassembled pkg)
  • support ip6 defragmenting
  • support sctp
  • support relative times in tags
  • add tests
  • make query language simpler (less @'s)
  • improve import speed by ignoring timedout packages instead of having to flush them before processing a new package
  • cache matching + uncertain streams for tags

web:

  • history reverse search with strg+r
  • improve/document graph ui
  • diffing of two streams
  • render http response in iframe with correct content type
  • add button to download raw data of a stream
  • autocomplete keywords while typing the query "nearley unparse"
  • show matching generated marks in stream view
  • let large tag queries and names overflow instead of widening the page layout
  • add open in CyberChef button to stream chunks
  • highlight data matches in stream view

both:

filters / converters

the converter feature will be implemented like this:

  • converters are executed serverside
    • each executable file in converters/ is considered a converter
      • an example is b64decode.py which decodes every received chunk using b64
    • the protocol uses stdin/stdout, one json object per line or an empty line
      • pkappa sends the following lines to the filter:
        • first: general stream information json
        • one line per data chunk containing a json with the dir(ection) and data(encoded in base64)
        • one empty line terminating the chunks
      • the filter responds with the following lines:
        • one line per output data chunk formatted identical to the ones coming from pkappa
        • one empty line terminating the chunks
        • one general stream information json
  • include matching tags/services/marks in general stream information json?
  • tags, marks and services can be triggers for a collection of filters if they have a low complexity
    • they must not match on filtered-data for now, also indirectly via other tags/marks/services
      • currently they cannot match any data filter. data.none: should be allowed.
  • whenever pkappa becomes aware of a stream matching a tag/mark/service that triggers a filter but the output of that filter for this stream is not yet cached, it will queue up a filtering processing
    • all matches are queued up whenever a tag update job finishes. this could be optimized to only queue new / updated matches
  • whenever pkappa becomes aware of a stream no longer matching any tag/mark/service that triggers a filter but there exists a cache for the output of the given filter for the stream, that cached info is invalidated
  • rerun the converter if a stream is updated through new pcaps
  • the stream request api will get a parameter for selecting the filter to apply, it will support auto, none, filter:
    • the mode auto is the default and will return the original stream data or the single cached filtered stream (if there is exactly one)
  • there will be one cache file per active filter with this format:
    • [u64 stream id] [u8 varint chunk sizes] [client data] [server data]
    • when the stream id is ~0, this is an unused slot
    • only save the output if it differs from the plaintext stream data to save space
  • the search will be modified this way:
    • [cs]data filters will search in all currently available filtered outputs as well as the unmodified stream content
    • there will be modifiers for these [cs]data filters that allow to specify which of the filtered outputs are searched, or to specify exactly one output that is used
      • The modifier looks like [cs]data.convertername:content
      • none is a reserved converter name and selects the plain unprocessed stream data
    • [cs]bytes filters will support specifying the converter modifier too
  • when a filter was evaluated tags and services might be re-evaluated when they contain [cs]data filters, thats why those tags/services may not be used as triggers
  • keep stderr and exit code in all cases. keep stderr if stderr not empty, but the process exited as expected?
  • show stderr of filters in UI
    • stderr can be fetched from /api/converters/stderr/[name]
  • use states in filter json protocol and display which state we're currently in in UI for debugging filter scripts
  • name filters transformations? converters? -> converters it is
  • allow to run any converter for any stream even if not attached to a stream in the stream view
    • this could be used to implement the "stream to pwntools or python requests" generators
    • should indicate if the converter is also attached to one of the tags matching the stream
  • allow converters to add (generated) tags to a stream
  • option to mark converter output "informative" and render it differently than client/server traffic
    • e.g. to render the pwntools script generator output in an easy to copy way without the "client sent" coloring
  • split chunk into sub-chunks with different content-types to e.g. render images inline