The Transport Layer Security (TLS) protocol (RFC 2246, [2]) includes
features to negotiate selection of a lossless data compression method
as part of the TLS Handshake Protocol and then to apply the algorithm
associated with the selected method as part of the TLS Record
Protocol. TLS defines one standard compression method,
CompressionMethod.null, which specifies that data exchanged via the
record protocol will not be compressed. Although this single
compression method helps ensure that TLS implementations are
interoperable, the lack of additional standard compression methods
has limited the ability to develop interoperative implementations
that include data compression.
Friend Informational [Page 2]
RFC 3943 TLS Compression Using LZS November 2004
TLS is used extensively to secure client-server connections on the
World Wide Web. Although these connections can often be
characterized as short-lived and exchanging relatively small amounts
of data, TLS is also being used in environments where connections can
be long-lived and the amount of data exchanged can extend into
thousands or millions of octets. For example, TLS is now
increasingly being used as an alternative Virtual Private Network
(VPN) connection. Compression services have long been associated
with IPSec and PPTP VPN connections, so extending compression
services to TLS VPN connections preserves the user experience for any
VPN connection. Compression within TLS is one way to help reduce the
bandwidth and latency requirements associated with exchanging large
amounts of data while preserving the security services provided by
TLS.
This document describes an additional compression method associated
with a lossless data compression algorithm for use with TLS. This
document specifies the application of Lempel-Ziv-Stac (LZS)
compression, a lossless compression algorithm, to TLS record
payloads. This specification also assumes a thorough understanding
of the TLS protocol [2].
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in BCP 14, RFC 2119 [1].
As described in section 6 of RFC 2246 [2], TLS is a stateful
protocol. Compression methods used with TLS can be either stateful
(the compressor maintains its state through all compressed records)
or stateless (the compressor compresses each record independently),
but there seems to be little known benefit in using a stateless
compression method within TLS. The LZS compression method described
in this document is stateful.
Compression algorithms can occasionally expand, rather than compress,
input data. The worst-case expansion factor of the LZS compression
method is only 12.5%. Thus, TLS records of 15K bytes can never
exceed the expansion limits described in section 6.2.2 of RFC 2246
[2]. If TLS records of 16K bytes expand to an amount greater than
17K bytes, then the uncompressed version of the TLS record must be
transmitted, as described below.
Friend Informational [Page 3]
RFC 3943 TLS Compression Using LZS November 2004
The LZS CompressionMethod is a 16-bit index and is negotiated as
described in RFC 2246 [2] and RFC 3749 [3]. The LZS
CompressionMethod is stored in the TLS Record Layer connection state
as described in RFC 2246 [2].
IANA has assigned 64 as compression method identifier for applying
LZS compression to TLS record payloads.
Sharing compression histories between or among more than one TLS
session may potentially cause information leakage between the TLS
sessions, as pathological compressed data can potentially reference
data prior to the beginning of the current record. LZS
implementations guard against this situation. However, to avoid this
potential threat, implementations supporting TLS compression MUST use
separate compression histories for each TLS session. This is not a
limitation of LZS compression but is an artifact for any compression
algorithm.
Furthermore, the LZS compression history (as well as any compression
history) contains plaintext. Specifically, the LZS history contains
the last 2K bytes of plaintext of the TLS session. Thus, when the
TLS session terminates, the implementation SHOULD treat the history
as it does any plaintext (e.g., free memory, overwrite contents).
Starting with a sliding window compression history, similar to LZ1
[8], a new, enhanced compression algorithm identified as LZS was
developed. The LZS algorithm is a general-purpose lossless
compression algorithm for use with a wide variety of data types. Its
encoding method is very efficient, providing compression for strings
as short as two octets in length.
The LZS algorithm uses a sliding window of 2,048 bytes. During
compression, redundant sequences of data are replaced with tokens
that represent those sequences. During decompression, the original
sequences are substituted for the tokens in such a way that the
original data is exactly recovered. LZS differs from lossy
compression algorithms, such as those often used for video
compression, that do not exactly reproduce the original data. The
details of LZS compression can be found in section 3.5 below.
Friend Informational [Page 4]
RFC 3943 TLS Compression Using LZS November 2004
This standard specifies "stateful" compression -- that is,
maintaining the compression history between records within a
particular TLS compression session. Within each separate compression
history, the LZS CompressionMethod can maintain compression history
information when compressing and decompressing record payloads.
Stateful compression provides a higher compression ratio to be
achieved on the data stream, as compared to stateless compression
(resetting the compression history between every record),
particularly for small records.
Stateful compression requires both a reliable link and sequenced
record delivery to ensure that all records can be decompressed in the
same order they were compressed. Since TLS and lower-layer protocols
provide reliable, sequenced record delivery, compression history
information MAY be maintained and exploited when the LZS
CompressionMethod is used.
Furthermore, there MUST be a separate LZS compression history
associated with each open TLS session. This not only provides
enhanced security (no potential information leakage between sessions
via a shared compression history), but also enables superior
compression ratio (bit bandwidth on the connection) across all open
TLS sessions with compression. A shared history would require
resetting the compression (and decompression) history when switching
between TLS sessions, and a single history implementation would
require resetting the compression (and decompression) history between
each record.
The sender MUST reset the compression history prior to compressing
the first TLS record of a TLS session after TLS handshake completes.
It is advantageous for the sender to maintain the compression history
for all subsequent records processed during the TLS session. This
results in the greatest compression ratio for a given data set. In
either case, this compression history MUST NOT be used for any other
open TLS session, to ensure privacy between TLS sessions.
The sender MUST "flush" the compressor each time it transmits a
compressed record. Flushing means that all data going into the
compressor is included in the output, i.e., no data is retained in
the hope of achieving better compression. Flushing ensures that each
compressed record payload can be decompressed completely. Flushing is
necessary to prevent a record's data from spilling over into a later
record. This is important for synchronizing compressed data with the
authenticated and encrypted data in a TLS record. Flushing is
handled automatically in most LZS implementations.
Friend Informational [Page 5]
RFC 3943 TLS Compression Using LZS November 2004
When the TLS session terminates, the implementation SHOULD dispose of
the memory resources associated with the related TLS compression
history. That is, the compression history SHOULD be handled as the
TLS key material is handled.
The LZS CompressionMethod also features "decompressing" uncompressed
data in order to maintain the history if the "compressed" data
actually expanded. The LZS CompressionMethod record format
facilitates identifying whether records contain compressed or
uncompressed data. The LZS decoding process accommodates
decompressing either compressed or uncompressed data.
Prior to compression, the uncompressed data (TLSPlaintext.fragment)
is composed of a plaintext TLS record. After compression, the
compressed data (TLSCompressed.fragment) is composed of an 8-bit
TLSComp header followed by the compressed (or uncompressed) data.
The format of the 8-bit Flags TLSComp field is as follows:
0 1 2 3 4 5 6 7
+-----+-----+-----+-----+-----+-----+-----+-----+
| Res | Res | Res | Res | Res | Res | RST | C/U |
+-----+-----+-----+-----+-----+-----+-----+-----+
Res-Reserved
Reserved for future use. MUST be set to zero. MUST be ignored by
the receiving node.
Friend Informational [Page 6]
RFC 3943 TLS Compression Using LZS November 2004
RST-Reset Compression History
The RST bit is used to inform the decompressing peer that the
compression history in this TLS session was reset prior to the
data contained in this TLS record being compressed. When the RST
bit is set to "1", a compression history reset is performed; when
RST is set to "0", a compression history reset is not performed.
This bit MUST be set to a value of "1" for the first compressed
TLS transmitted record of a TLS session. This bit may also be
used by the transmitter for other exception cases when the
compression history must be reset.
C/U-Compressed/Uncompressed Bit
The C/U indicates whether the data field contains compressed or
uncompressed data. A value of 1 indicates compressed data (often
referred to as a compressed record), and a value of 0 indicates
uncompressed data (or an uncompressed record).
The LZS compression method, encoding format, and application examples
are described in RFC 1967 [6], RFC 1974 [5], and RFC 2395 [4].
Some implementations of LZS allow the sending compressor to select
from among several options to provide varying compression ratios,
processing speeds, and memory requirements. Other implementations of
LZS provide optimal compression ratio at byte-per-clock speeds.
The receiving LZS decompressor automatically adjusts to the settings
selected by the sender. Also, receiving LZS decompressors will
update the decompression history with uncompressed data. This
facilitates never obtaining less than a 1:1 compression ratio in the
session and never transmitting with expanded data.
The input to the payload compression algorithm is TLSPlaintext data
destined to an active TLS session with compression negotiated. The
output of the algorithm is a new (and hopefully smaller)
TLSCompressed record. The output payload contains the input
payload's data in either compressed or uncompressed format. The
input and output payloads are each an integral number of bytes in
length.
The output payload is always prepended with the TLSComp header. If
the uncompressed form is used, the output payload is identical to the
input payload, and the TLSComp header reflects uncompressed data.
Friend Informational [Page 7]
RFC 3943 TLS Compression Using LZS November 2004
If the compressed form is used, encoded as defined in ANSI X3.241
[7], and the TLSComp header reflects compressed data. The LZS
encoded format is repeated here for informational purposes ONLY.
<Compressed Stream> := [<Compressed String>*] <End Marker>
<Compressed String> := 0 <Raw Byte> | 1 <Compressed Bytes>
<Raw Byte> := <b><b><b><b><b><b><b><b> (8-bit byte)
<Compressed Bytes> := <Offset> <Length>
<Offset> := 1 <b><b><b><b><b><b><b> | (7-bit offset)
0 <b><b><b><b><b><b><b><b><b><b><b> (11-bit offset)
<End Marker> := 110000000
<b> := 1 | 0
<Length> :=
00 = 2 1111 0110 = 14
01 = 3 1111 0111 = 15
10 = 4 1111 1000 = 16
1100 = 5 1111 1001 = 17
1101 = 6 1111 1010 = 18
1110 = 7 1111 1011 = 19
1111 0000 = 8 1111 1100 = 20
1111 0001 = 9 1111 1101 = 21
1111 0010 = 10 1111 1110 = 22
1111 0011 = 11 1111 1111 0000 = 23
1111 0100 = 12 1111 1111 0001 = 24
1111 0101 = 13 ...
A datagram payload compressed with LZS always ends with the last
compressed data byte (also known as the <end marker>), which is used
to disambiguate padding. This allows trailing bits, as well as
bytes, to be considered padding.
The size of a compressed payload MUST be in whole octet units.
All TLS records processed with a TLS session state that includes LZS
compression are processed as follows. The reliable and efficient
transport of LZS compressed records in the TLS session depends on the
following processes.
Friend Informational [Page 8]
RFC 3943 TLS Compression Using LZS November 2004
The compression operation results in either compressed or
uncompressed data. When a TLS record is received, it is assigned to
a particular TLS context that includes the LZS compression history
buffer. It is processed according to ANSI X3.241-1994 to form
compressed data or used as is to form uncompressed data. For the
first record of the session, or for exception conditions, the
compression history MUST be cleared. In performing the compression
operation, the compression history MUST be updated when either a
compressed record or an uncompressed record is produced.
Uncompressed TLS records MAY be sent at any time. Uncompressed TLS
records MUST be sent if compression causes enough expansion to make
the data compression TLS record size exceed the MTU defined in
section 6.2.2 in RFC 2246. The output of the compression operation
is placed in the fragment field of the TLSCompressed structure
(TLSCompressed.fragment).
The TLSComp header byte is located just prior to the first byte of
the compressed TLS record in TLSCompressed.fragment. The C/U bit in
the TLSComp header is set according to whether the data field
contains compressed or uncompressed data. The RST bit in the TLSComp
header is set to "1" if the compression history was reset prior to
compressing the TLSplaintext.fragment that is composed of a
TLSCompressed.fragment. Uncompressed data MUST be transmitted (and
the C/U bit set to 0) if the "compressed" (expanded) data exceeded
17K bytes.
Prior to decompressing the first compressed TLS record in the TLS
session, the receiver MUST reset the decompression history.
Subsequent records are decompressed in the order received. The
receiver decompresses the Payload Data field according to the
encoding specified in section 3.5 above.
If the received datagram is not compressed, the receiver does not
need to perform decompression processing, and the Payload Data field
of the datagram is ready for processing by the next protocol layer.
After a TLS record is received from the peer and decrypted, the RST
and C/U bits MUST be checked.
If the C/U bit is set to "1", the resulting compressed data block
MUST be decompressed according to section 3.5 above.
If the C/U bit is set to "0", the specified decompression history
MUST be updated with the received uncompressed data.
Friend Informational [Page 9]
RFC 3943 TLS Compression Using LZS November 2004
If the RST bit is set to "1", the receiving decompression history MAY
be reset to an initial state prior to decompressing the TLS record.
(However, due to the characteristics of the Hifn LZS algorithm, a
decompression history reset is not required). After reset, any
compressed or uncompressed data contained in the record is processed.
During compression, there are two workable options for handling
records that expand:
1) Send the expanded data (as long as TLSCompressed.length is 17K or
less) and maintain the history, thus allowing loss of current
bandwidth but preserving future bandwidth on the link.
2) Send the uncompressed data and do not clear the compression
history; the decompressor will update its history, thus conserving
the current bandwidth and future bandwidth on the link.
The second option is the preferred option and SHOULD be implemented.
There is a third option:
3) Send the uncompressed data and clear the history, thus conserving
current bandwidth but allowing possible loss of future bandwidth
on the link.
This option SHOULD NOT be implemented.
The compression method identifiers specified in this document are
machine-readable numbers. As such, issues of human
internationalization and localization are not introduced.
Section 2 of RFC 3749 [3] describes a registry of compression method
identifiers to be maintained by the IANA and to be assigned within
three zones.
IANA has assigned an identifier for the LZS compression method from
the RFC 2434 Specification Required IANA pool, as described in
sections 2 and 5 of RFC 3749 [3].
The IANA-assigned compression method identifier for LZS is 64 decimal
(0x40).
Friend Informational [Page 10]
RFC 3943 TLS Compression Using LZS November 2004
This document does not introduce any topics that alter the threat
model addressed by TLS. The security considerations described
throughout RFC 2246 [2] apply here as well.
However, combining compression with encryption can sometimes reveal
information that would not have been revealed without compression.
Data that is the same length before compression might be a different
length after compression, so adversaries that observe the length of
the compressed data might be able to derive information about the
corresponding uncompressed data. Some symmetric encryption
ciphersuites do not hide the length of symmetrically encrypted data
at all. Others hide it to some extent but not fully. For example,
ciphersuites that use stream cipher encryption without padding do not
hide length at all; ciphersuites that use Cipher Block Chaining (CBC)
encryption with padding provide some length hiding, depending on how
the amount of padding is chosen. Use of TLS compression SHOULD take
into account that the length of compressed data may leak more
information than the length of the original uncompressed data.
Another security issue to be aware of is that the LZS compression
history contains plaintext. In order to prevent any kind of
information leakage outside the system, when a TLS session with
compression terminates, the implementation SHOULD treat the
compression history as it does plaintext -- that is, care should be
taken not to reveal the compression history in any form or to use it
again. This is described in sections 2.2 and 3.2 above.
This information leakage concept can be extended to the situation of
sharing a single compression history across more than one TLS
session, as addressed in section 2.2 above.
Other security issues are discussed in RFC 3749 [3].
The concepts described in this document were derived from RFC 1967
[6], RFC 1974 [5], RFC 2395 [4], and RFC 3749 [3]. The author
acknowledges the contributions of Scott Hollenbeck, Douglas Whiting,
and Russell Dietz, and help from Steve Bellovin, Russ Housley, and
Eric Rescorla.
Friend Informational [Page 11]
RFC 3943 TLS Compression Using LZS November 2004
[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.
[2] Dierks, T. and C. Allen, "The TLS Protocol Version 1.0", RFC
2246, January 1999.
[3] Hollenbeck, S. "Transport Layer Security Protocol Compression
Methods", RFC 3749, May 2004.
[4] Friend, R. and R. Monsour, "IP Payload Compression Using LZS",
RFC 2395, December 1998.
[5] Friend, R. and W. Simpson, "PPP Stac LZS Compression Protocol",
RFC 1974, August 1996.
[6] Schneider, K. and R. Friend, "PPP LZS-DCP Compression Protocol
(LZS-DCP)", RFC 1967, August 1996.
[7] American National Standards Institute, Inc., "Data Compression
Method for Information Systems," ANSI X3.241-1994, August 1994.
[8] Lempel, A. and J. Ziv, "A Universal Algorithm for Sequential
Data Compression", IEEE Transactions On Information Theory, Vol.
IT-23, No. 3, September 1977.
Author's Address
Robert Friend
Hifn
5973 Avenida Encinas
Carlsbad, CA 92008
US
EMail: rfriend@hifn.com
Friend Informational [Page 12]
RFC 3943 TLS Compression Using LZS November 2004
Full Copyright Statement
Copyright (C) The Internet Society (2004).
This document is subject to the rights, licenses and restrictions
contained in BCP 78, and at www.rfc-editor.org, and except as set
forth therein, the authors retain all their rights.
This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the ISOC's procedures with respect to rights in ISOC Documents can
be found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at ietf-
ipr@ietf.org.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
Friend Informational [Page 13]