Considerations around
Transport Header Confidentiality, Network Operations, and the Evolution of
Internet Transport Protocols
University of Aberdeen
Department of Engineering
Fraser Noble Building
Aberdeen, Scotland
AB24 3UE
United Kingdom
gorry@erg.abdn.ac.uk
http://www.erg.abdn.ac.uk/
University of Glasgow
School of Computing Science
Glasgow, Scotland
G12 8QQ
United Kingdom
csp@csperkins.org
https://csperkins.org/
Transport
TSVWG
transport design
operations and management
To protect user data and privacy, Internet transport protocols have
supported payload encryption and authentication for some time. Such
encryption and authentication are now also starting to be applied to the
transport protocol headers. This helps avoid transport protocol
ossification by middleboxes, mitigate attacks against the transport
protocol, and protect metadata about the communication. Current
operational practice in some networks inspect transport header
information within the network, but this is no longer possible when
those transport headers are encrypted.
This document discusses the possible impact when network traffic uses
a protocol with an encrypted transport header. It suggests issues to
consider when designing new transport protocols or features.
Introduction
The transport layer supports the end-to-end flow of data across a
network path, providing features such as connection establishment,
reliability, framing, ordering, congestion control, flow control, etc.,
as needed to support applications. One of the core functions of an
Internet transport is to discover and adapt to the characteristics of
the network path that is currently being used.
For some years, it has been common for the transport-layer payload to
be protected by encryption and authentication but for the transport-layer
headers to be sent unprotected. Examples of protocols that behave
in this manner include Transport Layer Security
(TLS) over TCP , Datagram TLS , the Secure
Real-time Transport Protocol , and tcpcrypt . The use of unencrypted transport headers has led some
network operators, researchers, and others to develop tools and
processes that rely on observations of transport headers both in
aggregate and at the flow level to infer details of the network's
behaviour and inform operational practice.
Transport protocols are now being developed that encrypt some or all
of the transport headers, in addition to the transport payload data. The
QUIC transport protocol
is an example of such a protocol. Such transport header encryption makes
it difficult to observe transport protocol behaviour from the vantage
point of the network. This document discusses some implications of
transport header encryption for network operators and researchers that
have previously observed transport headers, and it highlights some issues
to consider for transport protocol designers.
As discussed in , the IETF has
concluded that Pervasive Monitoring (PM) is a technical attack that
needs to be mitigated in the design of IETF protocols. This document
supports that conclusion. It also recognises that
states, "Making networks unmanageable to mitigate PM is not an acceptable outcome, but
ignoring PM would go against the consensus documented here. An
appropriate balance will emerge over time as real instances of this
tension are considered." This document is written to provide input to
the discussion around what is an appropriate balance by highlighting
some implications of transport header encryption.
Current uses of transport header information by network devices on
the Internet path are explained. These uses can be beneficial or
malicious. This is written to provide input to the discussion around
what is an appropriate balance by highlighting some implications of
transport header encryption.
Current Uses of Transport Headers within the Network
In response to pervasive surveillance
revelations and the IETF consensus that "Pervasive Monitoring Is an
Attack" , efforts are underway to increase
encryption of Internet traffic. Applying confidentiality to transport
header fields can improve privacy and can help to mitigate certain
attacks or manipulation of packets by devices on the network path, but
it can also affect network operations and measurement .
When considering what parts of the transport headers should be
encrypted to provide confidentiality and what parts should be visible
to network devices (including unencrypted but authenticated headers),
it is necessary to consider both the impact on network operations and
management and the implications for ossification and user privacy . Different parties will view the relative
importance of these concerns differently. For some, the benefits of
encrypting all the transport headers outweigh the impact of doing so;
others might analyse the security, privacy, and ossification impacts and
arrive at a different trade-off.
This section reviews examples of the observation of transport-layer
headers within the network by using devices on the network path or by using
information exported by an on-path device. Unencrypted transport headers
provide information that can support network operations and management,
and this section notes some ways in which this has been done.
Unencrypted transport header information also contributes metadata that
can be exploited for purposes unrelated to network transport
measurement, diagnostics, or troubleshooting (e.g., to block or to
throttle traffic from a specific content provider), and this section
also notes some threats relating to unencrypted transport headers.
Exposed transport information also provides a source of information
that contributes to linked data sets, which could be exploited to deduce
private information, e.g., user patterns, user location, tracking
behaviour, etc. This might reveal information the parties did not intend
to be revealed. aims to make designers,
implementers, and users of Internet protocols aware of privacy-related
design choices in IETF protocols.
This section does not consider intentional modification of transport
headers by middleboxes, such as devices performing Network Address
Translation (NAT) or firewalls.
To Separate Flows in Network Devices
Some network-layer mechanisms separate network traffic by flow
without resorting to identifying the type of traffic: hash-based
load sharing across paths (e.g., Equal-Cost Multipath
(ECMP)); sharing across a group of links (e.g., using a Link Aggregation
Group (LAG)); ensuring equal access to link capacity (e.g., Fair
Queuing (FQ)); or distributing traffic to servers (e.g., load
balancing). To prevent packet reordering, forwarding engines can
consistently forward the same transport flows along the same
forwarding path, often achieved by calculating a hash using an n-tuple
gleaned from a combination of link header information through to
transport header information. This n-tuple can use the Media Access Control
(MAC) address and IP
addresses and can include observable transport header information.
When transport header information cannot be observed, there can be
less information to separate flows at equipment along the path.
Flow
separation might not be possible when a transport forms traffic
into an encrypted aggregate. For IPv6, the Flow Label can be used even when all transport
information is encrypted, enabling Flow Label-based ECMP and load sharing .
To Identify Transport Protocols and Flows
Information in exposed transport-layer headers can be used by the
network to identify transport protocols and flows . The ability to identify transport protocols,
flows, and sessions is a common function performed, for example, by
measurement activities, Quality of Service (QoS) classifiers, and
firewalls. These functions can be beneficial and performed with the
consent of, and in support of, the end user. Alternatively, the same
mechanisms could be used to support practises that might be
adversarial to the end user, including blocking, deprioritising, and
monitoring traffic without consent.
Observable transport header information, together with information
in the network header, has been used to identify flows and their
connection state, together with the set of protocol options being
used. Transport protocols, such as TCP
and the Stream Control Transmission Protocol (SCTP) , specify a standard base header that includes
sequence number information and other data. They also have the
possibility to negotiate additional headers at connection setup,
identified by an option number in the transport header.
In some uses, an assigned transport port (e.g., 0..49151) can
identify the upper-layer protocol or service . However, port information alone is not
sufficient to guarantee identification. Applications can use arbitrary
ports and do not need to use assigned port numbers. The use of an
assigned port number is also not limited to the protocol for which the
port is intended. Multiple sessions can also be multiplexed on a
single port, and ports can be reused by subsequent sessions.
Some flows can be identified by observing signalling data
(e.g., see and ) or
through the use of magic numbers placed in the first byte(s) of a
datagram payload .
When transport header information cannot be observed, this removes
information that could have been used to classify flows by passive
observers along the path. More ambitious ways could be used to
collect, estimate, or infer flow information, including heuristics
based on the analysis of traffic patterns, such as classification of
flows relying on timing, volumes of information, and correlation
between multiple flows. For example, an operator that cannot access
the Session Description Protocol (SDP) session descriptions to classify a flow as audio traffic might
instead use (possibly less-reliable) heuristics to infer that short
UDP packets with regular spacing carry audio traffic. Operational
practises aimed at inferring transport parameters are out of scope for
this document, and are only mentioned here to recognise that
encryption does not prevent operators from attempting to apply
practises that were used with unencrypted transport headers.
The IAB has provided a summary of
expected implications of increased encryption on network functions
that use the observable headers and describe the expected benefits of
designs that explicitly declare protocol-invariant header information
that can be used for this purpose.
To Understand Transport Protocol Performance
This subsection describes use by the network of exposed transport-layer headers to
understand transport protocol performance and
behaviour.
Using Information Derived from Transport-Layer Headers
Observable transport headers enable explicit measurement and
analysis of protocol performance and detection of network anomalies
at any point along the Internet path. Some operators use passive
monitoring to manage their portion of the Internet by characterising
the performance of link/network segments. Inferences from transport
headers are used to derive performance metrics:
- Traffic Rate and Volume:
- Per-application traffic
rate and volume measures can be used to characterise the traffic
that uses a network segment or the pattern of network usage.
Observing the protocol sequence number and packet size offers
one way to measure this (e.g., measurements observing counters
in periodic reports, such as RTCP , or measurements observing
protocol sequence numbers in statistical samples of packet
flows or specific control packets, such as those observed at
the start and end of a flow).
Measurements can be per endpoint or for an
endpoint aggregate. These could be used to assess usage or for
subscriber billing.
Such measurements can be used to trigger traffic
shaping and to associate QoS support within the network and
lower layers. This can be done with consent and in support of an
end user to improve quality of service or could be used by the
network to deprioritise certain flows without user consent.
The traffic rate and volume can be determined,
providing that the packets belonging to individual flows can be
identified, but there might be no additional information about a
flow when the transport headers cannot be observed.
- Loss Rate and Loss Pattern:
- Flow loss rate can be
derived (e.g., from transport sequence numbers or inferred from
observing transport protocol interactions) and has been used as
a metric for performance assessment and to characterise
transport behaviour. Network operators have used the variation
in patterns to detect changes in the offered service.
Understanding the location and root cause of loss can help an
operator determine whether this requires corrective action.
There are various causes of loss, including: corruption of
link frames (e.g., due to interference on a radio link);
buffering loss (e.g., overflow due to congestion, Active Queue
Management (AQM) , or inadequate
provision following traffic preemption), and policing (e.g., traffic
management ). Understanding flow
loss rates requires maintaining the per-flow state (flow
identification often requires transport-layer information) and
either observing the increase in sequence numbers in the network
or transport headers or comparing a per-flow packet counter
with the number of packets that the flow actually sent. Per-hop
loss can also sometimes be monitored at the interface level by
devices on the network path or by using in-situ methods operating
over a network segment (see ).
The pattern of loss can provide insight into the cause of
loss. Losses can often occur as bursts, randomly timed events,
etc. It can also be valuable to understand the conditions under
which loss occurs. This usually requires relating loss to the
traffic flowing at a network node or segment at the time of
loss. Transport header information can help identify cases where
loss could have been wrongly identified or where the transport
did not require retransmission of a lost packet.
- Throughput and Goodput:
- Throughput is the amount
of payload data sent by a flow per time interval. Goodput (the
subset of throughput consisting of useful traffic; see and ) is
a measure of useful data exchanged.
The throughput of a flow can be determined in the absence of
transport header information, providing that the individual flow
can be identified, and the overhead known. Goodput requires the
ability to differentiate loss and retransmission of packets, for
example, by observing packet sequence numbers in the TCP or RTP
headers .
- Latency:
- Latency is a key performance metric that
impacts application and user-perceived response times. It often
indirectly impacts throughput and flow completion time. This
determines the reaction time of the transport protocol itself,
impacting flow setup, congestion control, loss recovery, and
other transport mechanisms. The observed latency can have many
components . Of these,
unnecessary/unwanted queueing in buffers of the network devices
on the path has often been observed as a significant factor
. Once the cause of unwanted
latency has been identified, this can often be eliminated.
To measure latency across a part of a path, an observation
point can measure the experienced
round-trip time (RTT) by using packet sequence numbers and
acknowledgements or by observing header timestamp information.
Such information allows an observation point on the network path
to determine not only the path RTT but also allows measurement
of the upstream and downstream contribution to the RTT. This
could be used to locate a source of latency, e.g., by observing
cases where the median RTT is much greater than the minimum RTT
for a part of a path.
The service offered by network operators can benefit from
latency information to understand the impact of configuration
changes and to tune deployed services. Latency metrics are key
to evaluating and deploying AQM ,
Diffserv , and
Explicit Congestion
Notification (ECN) . Measurements could identify
excessively large buffers, indicating where to deploy or
configure AQM. An AQM method is often deployed in combination
with other techniques, such as scheduling
, and
although parameter-less methods are desired
, current methods often require tuning
because they cannot scale across
all possible deployment scenarios.
Latency and round-trip time information can potentially
expose some information useful for approximate geolocation, as
discussed in .
- Variation in Delay:
- Some network applications are
sensitive to (small) changes in packet timing (jitter). Short-
and long-term delay variation can impact the latency of a
flow and hence the perceived quality of applications using a
network path. For example, jitter metrics are often cited when
characterising paths supporting real-time traffic. The expected
performance of such applications can be inferred from a measure
of the variation in delay observed along a portion of the path
.
The requirements resemble those for the measurement of
latency.
- Flow Reordering:
- Significant packet reordering
within a flow can impact time-critical applications and can be
interpreted as loss by reliable transports. Many transport
protocol techniques are impacted by reordering (e.g., triggering
TCP retransmission or rebuffering of real-time applications).
Packet reordering can occur for many reasons, e.g., from equipment
design to misconfiguration of forwarding rules. Flow
identification is often required to avoid significant packet
misordering (e.g., when using ECMP, or LAG). Network tools can
detect and measure unwanted/excessive reordering and the impact
on transport performance.
There have been initiatives in the IETF transport area to
reduce the impact of reordering within a transport flow,
possibly leading to a reduction in the requirements for
preserving ordering. These have potential to simplify network
equipment design as well as the potential to improve robustness
of the transport service. Measurements of reordering can help
understand the present level of reordering and inform decisions
about how to progress new mechanisms.
Techniques for measuring reordering typically observe packet
sequence numbers. Metrics have been defined that evaluate
whether a network path has maintained packet order on a
packet-by-packet basis . Some protocols provide in-built
monitoring and reporting functions. Transport fields in the RTP
header can be observed to derive traffic
volume measurements and provide information on the progress and
quality of a session using RTP. Metadata assists in
understanding the context under which the data was collected,
including the time, observation point , and
way in which metrics were
accumulated. The RTCP protocol directly reports some of this
information in a form that can be directly visible by devices on
the network path.
In some cases, measurements could involve active injection of
test traffic to perform a measurement (see ). However, most operators do not have
access to user equipment; therefore, the point of test is normally
different from the transport endpoint. Injection of test traffic can
incur an additional cost in running such tests (e.g., the
implications of capacity tests in a mobile network segment are
obvious). Some active measurements
(e.g., response under load or particular workloads) perturb other
traffic and could require dedicated access to the network
segment.
Passive measurements (see )
can have advantages in terms of
eliminating unproductive test traffic, reducing the influence of
test traffic on the overall traffic mix, and having the ability to choose
the point of observation (see ).
Measurements can rely on observing packet headers, which is not
possible if those headers are encrypted, but could utilise
information about traffic volumes or patterns of interaction to
deduce metrics.
Passive packet sampling techniques are also often used to scale
the processing involved in observing packets on high-rate links.
This exports only the packet header information of (randomly)
selected packets. Interpretation of the exported information relies
on understanding of the header information. The utility of these
measurements depends on the type of network segment/link and number
of mechanisms used by the network devices. Simple routers are
relatively easy to manage, but a device with more complexity demands
understanding of the choice of many system parameters.
Using Information Derived from Network-Layer Header Fields
Information from the transport header can be used by a
multi-field (MF) classifier as a part of policy framework. Policies
are commonly used for management of the QoS or Quality of Experience
(QoE) in resource-constrained networks or by firewalls to implement
access rules (see also ).
Policies can support user
applications/services or protect against unwanted or lower-priority
traffic ().
Transport-layer information can also be explicitly carried in
network-layer header fields that are not encrypted, serving as a
replacement/addition to the exposed transport header information
. This information can enable a
different forwarding treatment by the devices forming the network
path, even when a transport employs encryption to protect other
header information.
On the one hand, the user of a transport that multiplexes
multiple subflows might want to obscure the presence and
characteristics of these subflows. On the other hand, an encrypted
transport could set the network-layer information to indicate the
presence of subflows and to reflect the service requirements of
individual subflows. There are several ways this could be done:
- IP Address:
- Applications normally expose the
endpoint addresses used in the forwarding decisions in network
devices. Address and other protocol information can be used by an
MF classifier to determine how traffic is treated and hence affects the quality of
experience for a flow. Common issues concerning IP address
sharing are described in .
- Using the IPv6 Network-Layer Flow Label:
- A number
of Standards Track and Best Current Practice RFCs (e.g., , , and ) encourage endpoints to set the IPv6
Flow Label field of the network-layer header.
As per , IPv6 source nodes "SHOULD assign each
unrelated transport connection and application data stream to a
new flow."
A multiplexing transport could choose
to use multiple flow labels to allow the network to
independently forward subflows. provides further
guidance on choosing a flow label value, stating these
"should be chosen such that their bits exhibit a high
degree of variability" and chosen so that "third
parties should be unlikely to be able to guess the next value
that a source of flow labels will choose."
Once set, a flow label can provide information
that can help inform network-layer queueing and forwarding,
including use with IPsec ,
Equal-Cost Multipath routing, and Link Aggregation .
The choice of how to assign a flow label needs to
avoid introducing linkages between flows that a network device
could not otherwise observe. Inappropriate use by the transport
can have privacy implications (e.g., assigning the same label to
two independent flows that ought not to be classified similarly).
- Using the Network-Layer Differentiated Services Code Point:
- Applications
can expose their delivery expectations to network devices by
setting the Differentiated Services Code Point (DSCP) field of
IPv4 and IPv6 packets . For
example, WebRTC applications identify different forwarding
treatments for individual subflows (audio vs. video) based on
the value of the DSCP field ). This provides
explicit information to inform network-layer queueing and
forwarding, rather than an operator inferring traffic
requirements from transport and application headers via a
multi-field classifier. Inappropriate use by the transport can
have privacy implications (e.g., assigning a different DSCP to a
subflow could assist in a network device discovering the traffic
pattern used by an application). The field is mutable, i.e.,
some network devices can be expected to change this field. Since
the DSCP value can impact the quality of experience for a flow,
observations of service performance have to consider this field
when a network path supports differentiated service
treatment.
- Using Explicit Congestion Notification:
- Explicit Congestion Notification (ECN)
is a transport mechanism that uses the
ECN field in the network-layer header. Use of ECN explicitly
informs the network layer that a transport is ECN capable and
requests ECN treatment of the flow. An ECN-capable transport can
offer benefits when used over a path with equipment that
implements an AQM method with Congestion Experienced (CE) marking of IP packets , since it can react to congestion
without also having to recover from lost packets.
ECN exposes the presence of congestion. The reception of
CE-marked packets can be used to estimate the level of incipient
congestion on the upstream portion of the path from the point of
observation ().
Interpreting the marking behaviour (i.e., assessing congestion
and diagnosing faults) requires context from the transport
layer, such as path RTT.
AQM and ECN offer a range of algorithms and configuration
options. Tools therefore have to be available to network
operators and researchers to understand the implication of
configuration choices and transport behaviour as the use of ECN
increases and new methods emerge
.
- Network-Layer Options:
- Network protocols can carry
optional headers (see ). These can
explicitly expose transport header information to on-path
devices operating at the network layer (as discussed further in
).
IPv4 has provisions
for optional header fields. IP routers can examine these headers
and are required to ignore IPv4 options that they do not
recognise. Many current paths include network devices that
forward packets that carry options on a slower processing path.
Some network devices (e.g., firewalls) can be (and are)
configured to drop these packets .
BCP 186 provides
guidance on how operators should treat IPv4 packets
that specify options.
IPv6 can encode optional network-layer
information in separate headers that may be placed between the
IPv6 header and the upper-layer header
(e.g., the IPv6 Alternate Marking
Method , which
can be used to measure packet loss and delay metrics). The
Hop-by-Hop Options header, when present, immediately follows the
IPv6 header. IPv6 permits this header to be examined by any node
along the path if explicitly configured .
Careful use of the network-layer features (e.g., extension
headers can; see ) help provide similar
information in the case where the network is unable to inspect
transport protocol headers.
To Support Network Operations
Some network operators make use of on-path observations of
transport headers to analyse the service offered to the users of a
network segment and inform operational practice and can help
detect and locate network problems.
gives an operator's perspective about such use.
When observable transport header information is not available,
those seeking an understanding of transport behaviour and dynamics
might learn to work without that information. Alternatively, they
might use more limited measurements combined with pattern inference
and other heuristics to infer network behaviour (see ). Operational practises aimed at
inferring transport parameters are out of scope for this document and
are only mentioned here to recognise that encryption does not
necessarily stop operators from attempting to apply practises that
have been used with unencrypted transport headers.
This section discusses topics concerning observation of transport
flows, with a focus on transport measurement.
Problem Location
Observations of transport header information can be used to
locate the source of problems or to assess the performance of a
network segment. Often issues can only be understood in the context
of the other flows that share a particular path, particular device
configuration, interface port, etc. A simple example is monitoring
of a network device that uses a scheduler or active queue management
technique , where it could be
desirable to understand whether the algorithms are correctly
controlling latency or if overload protection is working. This
implies knowledge of how traffic is assigned to any subqueues used
for flow scheduling but can require information about how the
traffic dynamics impact active queue management, starvation
prevention mechanisms, and circuit breakers.
Sometimes correlating observations of headers at multiple points
along the path (e.g., at the ingress and egress of a network
segment) allows an observer to determine the contribution of a
portion of the path to an observed metric (e.g., to locate a source
of delay, jitter, loss, reordering, or congestion marking).
Network Planning and Provisioning
Traffic rate and volume measurements are used to help plan
deployment of new equipment and configuration in networks. Data is
also valuable to equipment vendors who want to understand traffic
trends and patterns of usage as inputs to decisions about planning
products and provisioning for new deployments.
Trends in aggregate traffic can be observed and can be related to
the endpoint addresses being used, but when transport header
information is not observable, it might be impossible to correlate
patterns in measurements with changes in transport protocols. This
increases the dependency on other indirect sources of information to
inform planning and provisioning.
Compliance with Congestion Control
The traffic that can be observed by on-path network devices (the
"wire image") is a function of transport protocol design/options,
network use, applications, and user characteristics. In general,
when only a small proportion of the traffic has a specific
(different) characteristic, such traffic seldom leads to operational
concern, although the ability to measure and monitor it is lower.
The desire to understand the traffic and protocol interactions
typically grows as the proportion of traffic increases. The
challenges increase when multiple instances of an evolving protocol
contribute to the traffic that share network capacity.
Operators can manage traffic load (e.g., when the network is
severely overloaded) by deploying rate limiters, traffic shaping, or
network transport circuit breakers .
The information provided by observing transport headers is a source
of data that can help to inform such mechanisms.
- Congestion Control Compliance of Traffic:
- Congestion control is a key transport function . Many network operators implicitly
accept that TCP traffic complies with a behaviour that is
acceptable for the shared Internet. TCP algorithms have been
continuously improved over decades and have reached a level of
efficiency and correctness that is difficult to match in custom
application-layer mechanisms .
A standards-compliant TCP stack provides congestion control
that is judged safe for use across the Internet. Applications
developed on top of well-designed transports can be expected to
appropriately control their network usage, reacting when the
network experiences congestion, by backing off and reducing the load
placed on the network. This is the normal expected behaviour for
IETF-specified transports (e.g., TCP and SCTP).
- Congestion Control Compliance for UDP Traffic:
- UDP
provides a minimal message-passing datagram transport that has
no inherent congestion control mechanisms. Because congestion
control is critical to the stable operation of the Internet,
applications and other protocols that choose to use UDP as a
transport have to employ mechanisms to prevent collapse, avoid
unacceptable contributions to jitter/latency, and establish
an acceptable share of capacity with concurrent traffic .
UDP flows that expose a well-known header can be observed to
gain understanding of the dynamics of a flow and its congestion
control behaviour. For example, tools exist to monitor various
aspects of RTP header information and RTCP reports for real-time
flows (see ). The Secure RTP and
RTCP extensions were explicitly
designed to expose some header information to enable such
observation while protecting the payload data.
A network operator can observe the headers of transport
protocols layered above UDP to understand if the datagram flows
comply with congestion control expectations. This can help
inform a decision on whether it might be appropriate to deploy
methods, such as rate limiters, to enforce acceptable usage. The
available information determines the level of precision with
which flows can be classified and the design space for
conditioning mechanisms (e.g., rate-limiting, circuit breaker
techniques , or blocking
uncharacterised traffic) .
When anomalies are detected, tools can interpret the transport
header information to help understand the impact of specific
transport protocols (or protocol mechanisms) on the other traffic
that shares a network. An observer on the network path can gain an
understanding of the dynamics of a flow and its congestion control
behaviour. Analysing observed flows can help to build confidence
that an application flow backs off its share of the network load
under persistent congestion and hence to understand whether the
behaviour is appropriate for sharing limited network capacity. For
example, it is common to visualise plots of TCP sequence numbers
versus time for a flow to understand how a flow shares available
capacity, deduce its dynamics in response to congestion, etc.
The ability to identify sources and flows that contribute to
persistent congestion is important to the safe operation of network
infrastructure and can inform configuration of network devices to
complement the endpoint congestion avoidance mechanisms to avoid a
portion of the network being driven into congestion collapse .
To Characterise "Unknown" Network Traffic
The patterns and types of traffic that share Internet capacity
change over time as networked applications, usage patterns, and
protocols continue to evolve.
Encryption can increase the volume of "unknown" or
"uncharacterised" traffic seen by the network. If these traffic
patterns form a small part of the traffic aggregate passing through
a network device or segment of the network path, the dynamics of the
uncharacterised traffic might not have a significant collateral
impact on the performance of other traffic that shares this network
segment. Once the proportion of this traffic increases, monitoring
the traffic can determine if appropriate safety measures have to be
put in place.
Tracking the impact of new mechanisms and protocols requires
traffic volume to be measured and new transport behaviours to be
identified. This is especially true of protocols operating over a
UDP substrate. The level and style of encryption needs to be
considered in determining how this activity is performed.
Traffic that cannot be classified typically receives a default
treatment. Some networks block or rate-limit traffic that cannot be
classified.
To Support Network Security Functions
On-path observation of the transport headers of packets can be
used for various security functions. For example, Denial of Service
(DoS) and Distributed DoS (DDoS) attacks against the infrastructure
or against an endpoint can be detected and mitigated by
characterising anomalous traffic (see ) on a shorter timescale. Other uses
include support for security audits (e.g., verifying the compliance
with cipher suites), client and application fingerprinting for
inventory, and alerts provided for network intrusion detection and
other next generation firewall functions.
When using an encrypted transport, endpoints can directly provide
information to support these security functions. Another method, if
the endpoints do not provide this information, is to use an on-path
network device that relies on pattern inferences in the traffic and
heuristics or machine learning instead of processing observed header
information. An endpoint could also explicitly cooperate with an
on-path device (e.g., a QUIC endpoint could share information about
current uses of connection IDs).
Network Diagnostics and Troubleshooting
Operators monitor the health of a network segment to support a
variety of operational tasks ,
including procedures to provide early warning and trigger action, e.g., to
diagnose network problems, to manage security threats (including
DoS), to evaluate equipment or protocol performance, or to respond
to user performance questions. Information about transport flows can
assist in setting buffer sizes and help identify whether
link/network tuning is effective. Information can also support
debugging and diagnosis of the root causes of faults that concern a
particular user's traffic and can support postmortem investigation
after an anomaly. Sections
and of provide further examples.
Network segments vary in their complexity. The design trade-offs
for radio networks are often very different from those of wired
networks . A radio-based network
(e.g., cellular mobile, enterprise Wireless LAN (WLAN), satellite
access/backhaul, point-to-point radio) adds a subsystem that
performs radio resource management, with impact on the available
capacity and potentially loss/reordering of packets. This impact
can differ by traffic type and can be correlated with link
propagation and interference. These can impact the cost and
performance of a provided service and is expected to increase in
importance as operators bring together heterogeneous types of
network equipment and deploy opportunistic methods to access a shared
radio spectrum.
Tooling and Network Operations
A variety of open source and proprietary tools have been deployed
that use the transport header information observable with widely
used protocols, such as TCP or RTP/UDP/IP. Tools that dissect network
traffic flows can alert to potential problems that are hard to
derive from volume measurements, link statistics, or device
measurements alone.
Any introduction of a new transport protocol, protocol feature,
or application might require changes to such tools and could
impact operational practice and policies. Such changes have
associated costs that are incurred by the network operators that
need to update their tooling or develop alternative practises that
work without access to the changed/removed information.
The use of encryption has the desirable effect of preventing
unintended observation of the payload data, and these tools seldom
seek to observe the payload or other application details. A flow
that hides its transport header information could imply "don't
touch" to some operators. This might limit a trouble-shooting
response to "can't help, no trouble found".
An alternative that does not require access to an observable
transport headers is to access endpoint diagnostic tools or to
include user involvement in diagnosing and troubleshooting unusual
use cases or to troubleshoot nontrivial problems. Another approach
is to use traffic pattern analysis. Such tools can provide useful
information during network anomalies (e.g., detecting significant
reordering, high or intermittent loss); however, indirect
measurements need to be carefully designed to provide information
for diagnostics and troubleshooting.
If new protocols, or protocol extensions, are made to closely
resemble or match existing mechanisms, then the changes to tooling
and the associated costs can be small. Equally, more extensive
changes to the transport tend to require more extensive, and more
expensive, changes to tooling and operational practice. Protocol
designers can mitigate these costs by explicitly choosing to expose
selected information as invariants that are guaranteed not to change
for a particular protocol (e.g., the header invariants and the
spin bit in QUIC ).
Specification of common log formats and development of alternative
approaches can also help mitigate the costs of transport
changes.
To Mitigate the Effects of Constrained Networks
Some link and network segments are constrained by the capacity they
can offer by the time it takes to access capacity (e.g., due to
underlying radio resource management methods) or by asymmetries in
the design (e.g., many link are designed so that the capacity
available is different in the forward and return directions; some
radio technologies have different access methods in the forward and
return directions resulting from differences in the power budget).
The impact of path constraints can be mitigated using a proxy
operating at or above the transport layer to use an alternate
transport protocol.
In many cases, one or both endpoints are unaware of the
characteristics of the constraining link or network segment, and
mitigations are applied below the transport layer. Packet
classification and QoS methods (described in various sections) can be
beneficial in differentially prioritising certain traffic when there
is a capacity constraint or additional delay in scheduling link
transmissions. Another common mitigation is to apply header
compression over the specific link or subnetwork (see ).
To Provide Header Compression
Header compression saves link capacity by compressing network and
transport protocol headers on a per-hop basis. This has been widely
used with low bandwidth dial-up access links and still finds
application on wireless links that are subject to capacity
constraints. These methods are effective for bit-congestive links
sending small packets (e.g., reducing the cost for sending control
packets or small data packets over radio links).
Examples of header compression include use with TCP/IP and
RTP/UDP/IP flows . Successful
compression depends on observing the transport headers and
understanding the way fields change between packets and is hence
incompatible with header encryption. Devices that compress transport
headers are dependent on a stable header format, implying
ossification of that format.
Introducing a new transport protocol, or changing the format of
the transport header information, will limit the effectiveness of
header compression until the network devices are updated. Encrypting
the transport protocol headers will tend to cause the header
compression to fall back to compressing only the network-layer
headers, with a significant reduction in efficiency. This can limit
connectivity if the resulting flow exceeds the link capacity or if
the packets are dropped because they exceed the link Maximum
Transmission Unit (MTU).
The Secure RTP (SRTP) extensions
were explicitly designed to leave the transport protocol headers
unencrypted, but authenticated, since support for header compression
was considered important.
To Verify SLA Compliance
Observable transport headers coupled with published transport
specifications allow operators and regulators to explore and verify
compliance with Service Level Agreements (SLAs). It can also be used
to understand whether a service is providing differential treatment to
certain flows.
When transport header information cannot be observed, other methods
have to be found to confirm that the traffic produced conforms to the
expectations of the operator or developer.
Independently verifiable performance metrics can be utilised to
demonstrate regulatory compliance in some jurisdictions and as a
basis for informing design decisions. This can bring assurance to
those operating networks, often avoiding deployment of complex
techniques that routinely monitor and manage Internet traffic flows
(e.g., avoiding the capital and operational costs of deploying flow
rate-limiting and network circuit breaker methods ).
Research, Development, and Deployment
Research and development of new protocols and mechanisms need to be
informed by measurement data (as described in the previous section).
Data can also help promote acceptance of proposed standards
specifications by the wider community (e.g., as a method to judge the
safety for Internet deployment).
Observed data is important to ensure the health of the research and
development communities and provides data needed to evaluate new
proposals for standardisation. Open standards motivate a desire to
include independent observation and evaluation of performance and
deployment data. Independent data helps compare different methods, judge
the level of deployment, and ensure the wider applicability of the
results. This is important when considering when a protocol or mechanism
should be standardised for use in the general Internet. This, in turn,
demands control/understanding about where and when measurement samples
are collected. This requires consideration of the methods used to
observe information and the appropriate balance between encrypting all
and no transport header information.
There can be performance and operational trade-offs in exposing
selected information to network tools. This section explores key
implications of tools and procedures that observe transport protocols
but does not endorse or condemn any specific practises.
Independent Measurement
Encrypting transport header information has implications on the way
network data is collected and analysed. Independent observations by
multiple actors is currently used by the transport community to
maintain an accurate understanding of the network within transport
area working groups, IRTF research groups, and the broader research
community. This is important to be able to provide accountability and
demonstrate that protocols behave as intended; although, when providing
or using such information, it is important to consider the privacy of
the user and their incentive for providing accurate and detailed
information.
Protocols that expose the state of the transport protocol in their
header (e.g., timestamps used to calculate the RTT, packet numbers
used to assess congestion, and requests for retransmission) provide an
incentive for a sending endpoint to provide consistent information,
because a protocol will not work otherwise. An on-path observer can
have confidence that well-known (and ossified) transport header
information represents the actual state of the endpoints when this
information is necessary for the protocol's correct operation.
Encryption of transport header information could reduce the range
of actors that can observe useful data. This would limit the
information sources available to the Internet community to understand
the operation of new transport protocols, reducing information to
inform design decisions and standardisation of the new protocols and
related operational practises. The cooperating dependence of network,
application, and host to provide communication performance on the
Internet is uncertain when only endpoints (i.e., at user devices and
within service platforms) can observe performance and when
performance cannot be independently verified by all parties.
Measurable Transport Protocols
Transport protocol evolution and the ability to measure and
understand the impact of protocol changes have to proceed
hand-in-hand. A transport protocol that provides observable headers
can be used to provide open and verifiable measurement data.
Observation of pathologies has a critical role in the design of
transport protocol mechanisms and development of new mechanisms and
protocols and aides in understanding the interactions between
cooperating protocols and network mechanisms, the implications of
sharing capacity with other traffic, and the impact of different
patterns of usage. The ability of other stakeholders to review
transport header traces helps develop insight into the performance and
the traffic contribution of specific variants of a protocol.
Development of new transport protocol mechanisms has to consider
the scale of deployment and the range of environments in which the
transport is used. Experience has shown that it is often difficult to
correctly implement new mechanisms and
that mechanisms often evolve as a protocol matures or in response to
changes in network conditions, in network traffic, or
to application usage. Analysis is especially valuable when based on
the behaviour experienced across a range of topologies, vendor
equipment, and traffic patterns.
Encryption enables a transport protocol to choose which internal
state to reveal to devices on the network path, what information to
encrypt, and what fields to grease . A
new design can provide summary information regarding its performance,
congestion control state, etc., or make explicit
measurement information available. For example,
specifies a way for a QUIC
endpoint to optionally set the spin bit to explicitly reveal the RTT
of an encrypted transport session to the on-path network devices.
There is a choice of what information to expose. For some operational
uses, the information has to contain sufficient detail to understand,
and possibly reconstruct, the network traffic pattern for further
testing. The interpretation of the information needs to consider
whether this information reflects the actual transport state of the
endpoints. This might require the trust of transport protocol
implementers to correctly reveal the desired information.
New transport protocol formats are expected to facilitate an
increased pace of transport evolution and with it the possibility to
experiment with and deploy a wide range of protocol mechanisms. At the
time of writing, there has been interest in a wide range of new
transport methods, e.g., larger initial window, Proportional Rate
Reduction (PRR), congestion control methods based on measuring
bottleneck bandwidth and round-trip propagation time, the introduction
of AQM techniques, and new forms of ECN response (e.g., Data Centre
TCP, DCTCP, and methods proposed for Low Latency Low Loss Scalable throughput (L4S)). The growth and diversity of
applications and protocols using the Internet also continues to
expand. For each new method or application, it is desirable to build a
body of data reflecting its behaviour under a wide range of deployment
scenarios, traffic load, and interactions with other
deployed/candidate methods.
Other Sources of Information
Some measurements that traditionally rely on observable transport
information could be completed by utilising endpoint-based logging
(e.g., based on QUIC trace and
qlog). Such information
has a diversity of uses, including developers wishing to
debug/understand the transport/application protocols with which they
work, researchers seeking to spot trends and anomalies, and
to characterise variants of protocols. A standard format for endpoint
logging could allow these to be shared (after appropriate
anonymisation) to understand performance and pathologies.
When measurement datasets are made available by servers or client
endpoints, additional metadata, such as the state of the network and
conditions in which the system was observed, is often necessary to
interpret this data to answer questions about network performance or
understand a pathology. Collecting and coordinating such metadata is
more difficult when the observation point is at a different location
to the bottleneck or device under evaluation .
Despite being applicable in some scenarios, endpoint logs do not
provide equivalent information to on-path measurements made by devices
in the network. In particular, endpoint logs contain only a part of
the information to understand the operation of network devices and
identify issues, such as link performance or capacity sharing between
multiple flows. An analysis can require coordination between actors at
different layers to successfully characterise flows and correlate the
performance or behaviour of a specific mechanism with an equipment
configuration and traffic using operational equipment along a network
path (e.g., combining transport and network measurements to explore
congestion control dynamics to understand the implications of traffic
on designs for active queue management or circuit breakers).
Another source of information could arise from Operations,
Administration, and Maintenance (OAM) (see ).
Information data records could be embedded into header information at
different layers to support functions, such as performance evaluation,
path tracing, path verification information, classification, and a
diversity of other uses.
In-situ OAM (IOAM) data fields can be encapsulated into a
variety of protocols to record operational and telemetry information
in an existing packet while that packet traverses a part of the path
between two points in a network (e.g., within a particular IOAM
management domain). IOAM-Data-Fields are independent from the
protocols into which IOAM-Data-Fields are encapsulated. For example, IOAM
can provide proof that a traffic flow takes a
predefined path, SLA verification for the live data traffic, and
statistics relating to traffic distribution.
Encryption and Authentication of Transport Headers
There are several motivations for transport header encryption.
One motive to encrypt transport headers is to prevent network
ossification from network devices that inspect well-known transport
headers. Once a network device observes a transport header and becomes
reliant upon using it, the overall use of that field can become
ossified, preventing new versions of the protocol and mechanisms from
being deployed. Examples include:
- During the development of TLS 1.3 ,
the design needed to function in the presence of deployed
middleboxes that relied on the presence of certain header fields
exposed in TLS 1.2 .
- The design of Multipath TCP (MPTCP) had to account for middleboxes (known as
"TCP Normalizers") that monitor the evolution of the window
advertised in the TCP header and then reset connections when the
window did not grow as expected.
- TCP Fast Open can experience
problems due to middleboxes that modify the transport header of
packets by removing "unknown" TCP options. Segments with
unrecognised TCP options can be dropped, segments that contain data
and set the SYN bit can be dropped, and some middleboxes that
disrupt connections can send data before completion of the
three-way handshake.
- Other examples of TCP ossification have included middleboxes that
modify transport headers by rewriting TCP sequence and
acknowledgement numbers but are unaware of the (newer) TCP
selective acknowledgement (SACK) option and therefore fail to
correctly rewrite the SACK information to match the changes made to
the fixed TCP header, preventing correct SACK operation.
In all these cases, middleboxes with a hard-coded, but incomplete,
understanding of a specific transport behaviour (i.e., TCP) interacted
poorly with transport protocols after the transport behaviour was
changed. In some cases, the middleboxes modified or replaced information
in the transport protocol header.
Transport header encryption prevents an on-path device from observing
the transport headers and therefore stops ossified mechanisms being
used that directly rely on or infer semantics of the transport header
information. This encryption is normally combined with authentication of
the protected information. summarises this
approach, stating
that "[t]he wire image, not the protocol's specification, determines
how third parties on the network paths among protocol participants will
interact with that protocol" (), and it can be expected that header information that is not
encrypted will become ossified.
Encryption does not itself prevent ossification of the network
service. People seeking to understand or classify network traffic could
still come to rely on pattern inferences and other heuristics or machine
learning to derive measurement data and as the basis for network
forwarding decisions . This can also
create dependencies on the transport protocol or the patterns of
traffic it can generate, also resulting in ossification of the
service.
Another motivation for using transport header encryption is to
improve privacy and to decrease opportunities for surveillance. Users
value the ability to protect their identity and location and defend
against analysis of the traffic. Revelations about the use of pervasive
surveillance have, to some extent, eroded
trust in the service offered by network operators and have led to an
increased use of encryption. Concerns have also been voiced about the
addition of metadata to packets by third parties to provide analytics,
customisation, advertising, cross-site tracking of users,
customer billing, or selectively allowing or blocking content.
Whatever the reasons, the IETF is designing protocols that include
transport header encryption (e.g., QUIC ) to supplement the already
widespread payload encryption and to further limit exposure of
transport metadata to the network.
If a transport protocol uses header encryption, the designers have to
decide whether to encrypt all or a part of the transport-layer
information. states,
"Anything exposed to the path should be done with the intent that it be
used by the network elements on the path."
Certain transport header fields can be made observable to on-path
network devices or can define new fields designed to explicitly expose
observable transport-layer information to the network. Where exposed
fields are intended to be immutable (i.e., can be observed but not
modified by a network device), the endpoints are encouraged to use
authentication to provide a cryptographic integrity check that can
detect if these immutable fields have been modified by network devices.
Authentication can help to prevent attacks that rely on sending packets
that fake exposed control signals in transport headers (e.g., TCP RST
spoofing). Making a part of a transport header observable or exposing
new header fields can lead to ossification of that part of a header as
network devices come to rely on observations of the exposed fields.
The use of transport header authentication and encryption therefore
exposes a tussle between middlebox vendors, operators, researchers,
applications developers, and end users:
- On the one hand, future Internet protocols that support transport
header encryption assist in the restoration of the end-to-end nature
of the Internet by returning complex processing to the endpoints.
Since middleboxes cannot modify what they cannot see, the use of
transport header encryption can improve application and end-user
privacy by reducing leakage of transport metadata to operators that
deploy middleboxes.
- On the other hand, encryption of transport-layer information has
implications for network operators and researchers seeking to
understand the dynamics of protocols and traffic patterns, since it
reduces the information that is available to them.
The following briefly reviews some security design options for
transport protocols. "A Survey of the Interaction between Security
Protocols and Transport Services" provides
more details concerning commonly used encryption methods at the
transport layer.
Security work typically employs a design technique that seeks to
expose only what is needed . This approach
provides incentives to not reveal any information that is not necessary
for the end-to-end communication. The IETF has provided guidelines for
writing security considerations for IETF specifications .
Endpoint design choices impacting privacy also need to be considered
as a part of the design process . The IAB
has provided guidance for analysing and documenting privacy
considerations within IETF specifications .
- Authenticating the Transport Protocol Header:
- Transport-layer header information can be authenticated. An example transport
authentication mechanism is TCP Authentication Option (TCP-AO) . This TCP option authenticates the IP
pseudo-header, TCP header, and TCP data. TCP-AO protects the
transport layer, preventing attacks from disabling the TCP
connection itself and provides replay protection. Such
authentication might interact with middleboxes, depending on their
behaviour .
The IPsec Authentication Header (AH)
was designed to work at the network layer and authenticate
the IP payload. This approach authenticates all transport headers
and verifies their integrity at the receiver, preventing
modification by network devices on the path. The IPsec Encapsulating
Security Payload (ESP) can also
provide authentication and integrity without confidentiality using
the NULL encryption algorithm . SRTP
is another example of a transport
protocol that allows header authentication.
- Integrity Check:
- Transport protocols usually employ
integrity checks on the transport header information. Security
methods usually employ stronger checks and can combine this with
authentication. An integrity check that protects the immutable
transport header fields, but can still expose the transport header
information in the clear, allows on-path network devices to observe
these fields. An integrity check is not able to prevent modification
by network devices on the path but can prevent a receiving endpoint
from accepting changes and avoid impact on the transport protocol
operation, including some types of attack.
- Selectively Encrypting Transport Headers and Payload:
- A
transport protocol design that encrypts selected header fields
allows specific transport header fields to be made observable by
network devices on the path. This information is explicitly exposed
either in a transport header field or lower layer protocol header. A
design that only exposes immutable fields can also perform
end-to-end authentication of these fields across the path to prevent
undetected modification of the immutable transport headers.
Mutable fields in the transport header provide opportunities
where on-path network devices can modify the transport behaviour
(e.g., the extended headers described in ). An example of a
method that encrypts some, but not all, transport header information
is GRE-in-UDP when used with GRE
encryption.
- Optional Encryption of Header Information:
- There are
implications to the use of optional header encryption in the design
of a transport protocol, where support of optional mechanisms can
increase the complexity of the protocol and its implementation and
in the management decisions that have to be made to use variable
format fields. Instead, fields of a specific type ought to be sent
with the same level of confidentiality or integrity protection.
- Greasing:
- Protocols often provide extensibility
features, reserving fields or values for use by future versions of a
specification. The specification of receivers has traditionally
ignored unspecified values; however, on-path network devices have
emerged that ossify to require a certain value in a field or reuse
a field for another purpose. When the specification is later
updated, it is impossible to deploy the new use of the field and
forwarding of the protocol could even become conditional on a
specific header field value.
A protocol can intentionally vary the value, format,
and/or presence of observable transport header fields at random
. This prevents a network device
ossifying the use of a specific observable field and can ease future
deployment of new uses of the value or code point. This is not a
security mechanism, although the use can be combined with an
authentication mechanism.
Different transports use encryption to protect their header
information to varying degrees. The trend is towards increased
protection.
Intentionally Exposing Transport Information to the Network
A transport protocol can choose to expose certain transport
information to on-path devices operating at the network layer by sending
observable fields. One approach is to make an explicit choice not to
encrypt certain transport header fields, making this transport
information observable by an on-path network device. Another approach is
to expose transport information in a network-layer extension header (see
). Both are examples of explicit information
intended to be used by network devices on the path .
Whatever the mechanism used to expose the information, a decision to
expose only specific information places the transport endpoint in
control of what to expose outside of the encrypted transport header.
This decision can then be made independently of the transport protocol
functionality. This can be done by exposing part of the transport header
or as a network-layer option/extension.
Exposing Transport Information in Extension Headers
At the network layer, packets can carry optional headers that
explicitly expose transport header information to the on-path devices
operating at the network layer (). For
example, an endpoint that sends an IPv6 hop-by-hop option can provide explicit transport-layer
information that can be observed and used by network devices on the
path. New hop-by-hop options are not recommended in "because nodes may be configured to
ignore the Hop-by-Hop Options header, drop packets containing a
Hop-by-Hop Options header, or assign packets containing a Hop-by-Hop
Options header to a slow processing path. Designers considering
defining new hop-by-hop options need to be aware of this likely
behavior."
Network-layer optional headers explicitly indicate the information
that is exposed, whereas use of exposed transport header information
first requires an observer to identify the transport protocol and its
format. See .
An arbitrary path can include one or more network devices that drop
packets that include a specific header or option used for this purpose
(see ). This could impact the proper
functioning of the protocols using the path. Protocol methods can be
designed to probe to discover whether the specific option(s) can be
used along the current path, enabling use on arbitrary paths.
Common Exposed Transport Information
There are opportunities for multiple transport protocols to
consistently supply common observable information . A common approach can result in an open
definition of the observable fields. This has the potential that the
same information can be utilised across a range of operational and
analysis tools.
Considerations for Exposing Transport Information
Considerations concerning what information, if any, it is
appropriate to expose include:
- On the one hand, explicitly exposing derived fields containing
relevant transport information (e.g., metrics for loss, latency,
etc.) can avoid network devices needing to derive this information
from other header fields. This could result in development and
evolution of transport-independent tools around a common
observable header and permit transport protocols to also evolve
independently of this ossified header .
- On the other hand, protocols and implementations might be
designed to avoid consistently exposing external information that
corresponds to the actual internal information used by the
protocol itself. An endpoint/protocol could choose to expose
transport header information to optimise the benefit it gets from
the network . The value of this
information for analysing operation of the transport layer would
be enhanced if the exposed information could be verified to match
the transport protocol's observed behavior.
The motivation to include actual transport header information and
the implications of network devices using this information has to be
considered when proposing such a method.
summarises this as:
When signals from endpoints to the path are independent from the
signals used by endpoints to manage the flow's state mechanics, they
may be falsified by an endpoint without affecting the peer's
understanding of the flow's state. For encrypted flows, this
divergence is not detectable by on-path devices.
Addition of Transport OAM Information to Network-Layer Headers
Even when the transport headers are encrypted, on-path devices can
make measurements by utilising additional protocol headers carrying OAM
information in an additional packet header. OAM information can be
included with packets to perform functions, such as identification of
transport protocols and flows, to aide understanding of network or
transport performance or to support network operations or mitigate the
effects of specific network segments.
Using network-layer approaches to reveal information has the
potential that the same method (and hence same observation and analysis
tools) can be consistently used by multiple transport protocols. This
approach also could be applied to methods beyond OAM (see ). There can also be less desirable implications
from separating the operation of the transport protocol from the
measurement framework.
Use of OAM within a Maintenance Domain
OAM information can be restricted to a maintenance domain,
typically owned and operated by a single entity. OAM information can
be added at the ingress to the maintenance domain (e.g., an Ethernet
protocol header with timestamps and sequence number information using
a method such as 802.11ag or in-situ OAM or as a part of the
encapsulation protocol). This additional header information is not
delivered to the endpoints and is typically removed at the egress of
the maintenance domain.
Although some types of measurements are supported, this approach
does not cover the entire range of measurements described in this
document. In some cases, it can be difficult to position measurement
tools at the appropriate segments/nodes, and there can be challenges in
correlating the downstream/upstream information when in-band OAM data
is inserted by an on-path device.
Use of OAM across Multiple Maintenance Domains
OAM information can also be added at the network layer by the
sender as an IPv6 extension header or an IPv4 option or in an
encapsulation/tunnel header that also includes an extension header or
option. This information can be used across multiple network segments
or between the transport endpoints.
One example is the IPv6 Performance and Diagnostic Metrics (PDM)
destination option . This allows a
sender to optionally include a destination option that carries header
fields that can be used to observe timestamps and packet sequence
numbers. This information could be authenticated by a receiving
transport endpoint when the information is added at the sender and
visible at the receiving endpoint, although methods to do this have
not currently been proposed. This needs to be explicitly enabled at
the sender.
Conclusions
Header authentication and encryption and strong integrity checks are being incorporated
into new transport protocols and have important benefits. The pace of the
development of transports using the WebRTC data channel and the rapid
deployment of the QUIC transport protocol can both be attributed to
using the combination of UDP as a substrate while providing
confidentiality and authentication of the encapsulated transport headers
and payload.
This document has described some current practises, and the
implications for some stakeholders, when transport-layer header
encryption is used. It does not judge whether these practises are
necessary or endorse the use of any specific practise. Rather, the
intent is to highlight operational tools and practises to consider when
designing and modifying transport protocols, so protocol designers can
make informed choices about what transport header fields to encrypt and
whether it might be beneficial to make an explicit choice to expose
certain fields to devices on the network path. In making such a
decision, it is important to balance:
- User Privacy:
- The less transport header information that is
exposed to the network, the lower the risk of leaking metadata that
might have user privacy implications. Transports that chose to
expose some header fields need to make a privacy assessment to
understand the privacy cost versus benefit trade-off in making that
information available. The design of the QUIC spin bit to the
network is an example of such considered analysis.
- Transport Ossification:
- Unencrypted transport header fields are
likely to ossify rapidly, as network devices come to rely on their
presence, making it difficult to change the transport in future.
This argues that the choice to expose information to the network is
made deliberately and with care, since it is essentially defining a
stable interface between the transport and the network. Some
protocols will want to make that interface as limited as possible;
other protocols might find value in exposing certain information to
signal to the network or in allowing the network to change certain
header fields as signals to the transport. The visible wire image of
a protocol should be explicitly designed.
- Network Ossification:
- While encryption can reduce ossification of
the transport protocol, it does not itself prevent ossification of
the network service. People seeking to understand network traffic
could still come to rely on pattern inferences and other heuristics
or machine learning to derive measurement data and as the basis for
network forwarding decisions . This
creates dependencies on the transport protocol or the patterns of
traffic it can generate, resulting in ossification of the
service.
- Impact on Operational Practice:
- The network operations community
has long relied on being able to understand Internet traffic
patterns, both in aggregate and at the flow level, to support
network management, traffic engineering, and troubleshooting.
Operational practice has developed based on the information
available from unencrypted transport headers. The IETF has supported
this practice by developing operations and management specifications, interface
specifications, and associated Best
Current Practices. Widespread deployment of transport protocols that
encrypt their information will impact network operations unless
operators can develop alternative practises that work without access
to the transport header.
- Pace of Evolution:
- Removing obstacles to change can enable an
increased pace of evolution. If a protocol changes its transport
header format (wire image) or its transport behaviour, this can
result in the currently deployed tools and methods becoming no
longer relevant. Where this needs to be accompanied by development
of appropriate operational support functions and procedures, it can
incur a cost in new tooling to catch up with each change. Protocols
that consistently expose observable data do not require such
development but can suffer from ossification and need to consider
if the exposed protocol metadata has privacy implications. There is
no single deployment context; therefore, designers need to
consider the diversity of operational networks (ISPs, enterprises,
DDoS mitigation and firewall maintainers, etc.).
- Supporting Common Specifications:
- Common, open, transport
specifications can stimulate engagement by developers, users,
researchers, and the broader community. Increased protocol diversity
can be beneficial in meeting new requirements, but the ability to
innovate without public scrutiny risks point solutions that optimise
for specific cases and that can accidentally disrupt operations
of/in different parts of the network. The social contract that
maintains the stability of the Internet relies on accepting common
transport specifications and on it being possible to detect
violations. The existence of independent measurements, transparency,
and public scrutiny of transport protocol behaviour helps the
community to enforce the social norm that protocol implementations
behave fairly and conform (at least mostly) to the specifications.
It is important to find new ways of maintaining that community trust
as increased use of transport header encryption limits visibility
into transport behaviour (see also ).
- Impact on Benchmarking and Understanding Feature Interactions:
- An appropriate vantage point for observation, coupled with timing
information about traffic flows, provides a valuable tool for
benchmarking network devices, endpoint stacks, and/or
configurations. This can help understand complex feature
interactions. An inability to observe transport header information
can make it harder to diagnose and explore interactions between
features at different protocol layers, a side effect of not allowing
a choice of vantage point from which this information is observed.
New approaches might have to be developed.
- Impact on Research and Development:
- Hiding transport header
information can impede independent research into new mechanisms,
measurements of behaviour, and development initiatives. Experience
shows that transport protocols are complicated to design and complex
to deploy and that individual mechanisms have to be evaluated while
considering other mechanisms across a broad range of network
topologies and with attention to the impact on traffic sharing the
capacity. If increased use of transport header encryption results in
reduced availability of open data, it could eliminate the
independent checks to the standardisation process that have
previously been in place from research and academic contributors
(e.g., the role of the IRTF Internet Congestion Control Research
Group (ICCRG) and research publications in reviewing new transport
mechanisms and assessing the impact of their deployment).
Observable transport header information might be useful to various
stakeholders. Other sets of stakeholders have incentives to limit what
can be observed. This document does not make recommendations about what
information ought to be exposed, to whom it ought to be observable, or
how this will be achieved. There are also design choices about where
observable fields are placed. For example, one location could be a part
of the transport header outside of the encryption envelope; another
alternative is to carry the information in a network-layer option or
extension header. New transport protocol designs ought to explicitly
identify any fields that are intended to be observed, consider if there
are alternative ways of providing the information, and reflect on the
implications of observable fields being used by on-path network devices
and how this might impact user privacy and protocol evolution when these
fields become ossified.
As notes, "Making networks
unmanageable to mitigate PM is not an acceptable
outcome, but ignoring PM would go against the
consensus documented here." Providing explicit information can help
avoid traffic being inappropriately classified, impacting application
performance. An appropriate balance will emerge over time as real
instances of this tension are analysed .
This balance between information exposed and information hidden ought to
be carefully considered when specifying new transport protocols.
Security Considerations
This document is about design and deployment considerations for
transport protocols. Issues relating to security are discussed
throughout this document.
Authentication, confidentiality protection, and integrity protection
are identified as transport features by .
As currently deployed in the Internet, these features are generally
provided by a protocol or layer on top of the transport protocol .
Confidentiality and strong integrity checks have properties that can
also be incorporated into the design of a transport protocol or to
modify an existing transport. Integrity checks can protect an endpoint
from undetected modification of protocol fields by on-path network
devices, whereas encryption and obfuscation or greasing can further
prevent these headers being utilised by network devices . Preventing observation of headers provides an
opportunity for greater freedom to update the protocols and can ease
experimentation with new techniques and their final deployment in
endpoints. A protocol specification needs to weigh the costs of
ossifying common headers versus the potential benefits of exposing
specific information that could be observed along the network path to
provide tools to manage new variants of protocols.
Header encryption can provide confidentiality of some or all of the
transport header information. This prevents an on-path device from
gaining knowledge of the header field. It therefore prevents mechanisms
being built that directly rely on the information or seeks to infer
semantics of an exposed header field. Reduced visibility into transport
metadata can limit the ability to measure and characterise traffic and
conversely can provide privacy benefits.
Extending the transport payload security context to also include the
transport protocol header protects both types of information with the
same key. A privacy concern would arise if this key was shared with a
third party, e.g., providing access to transport header information to
debug a performance issue would also result in exposing the transport
payload data to the same third party. Such risks would be mitigated
using a layered security design that provides one domain of protection
and associated keys for the transport payload and encrypted transport
headers and a separate domain of protection and associated keys for any
observable transport header fields.
Exposed transport headers are sometimes utilised as a part of the
information to detect anomalies in network traffic. As stated in , "While PM is an
attack, other forms of monitoring that might fit the definition of PM
can be beneficial and not part of any attack, e.g., network management
functions monitor packets or flows and anti-spam mechanisms need to see
mail message content." This can be used
as the first line of defence to identify potential threats from DoS or
malware and redirect suspect traffic to dedicated nodes responsible for
DoS analysis, for malware detection, or to perform packet "scrubbing" (the
normalisation of packets so that there are no ambiguities in
interpretation by the ultimate destination of the packet). These
techniques are currently used by some operators to also defend from
distributed DoS attacks.
Exposed transport header fields can also form a part of the
information used by the receiver of a transport protocol to protect the
transport layer from data injection by an attacker. In evaluating this
use of exposed header information, it is important to consider whether
it introduces a significant DoS threat. For example, an attacker could
construct a DoS attack by sending packets with a sequence number that
falls within the currently accepted range of sequence numbers at the
receiving endpoint. This would then introduce additional work at the
receiving endpoint, even though the data in the attacking packet might
not finally be delivered by the transport layer. This is sometimes known
as a "shadowing attack". An attack can, for example, disrupt
receiver processing, trigger loss and retransmission, or make a
receiving endpoint perform unproductive decryption of packets that
cannot be successfully decrypted (forcing a receiver to commit
decryption resources, or to update and then restore protocol state).
One mitigation to off-path attacks is to deny knowledge of what header
information is accepted by a receiver or obfuscate the accepted header
information, e.g., setting a nonpredictable initial value for a
sequence number during a protocol handshake, as in
and , or a port
value that cannot be predicted (see ). A receiver could also require additional
information to be used as a part of a validation check before accepting
packets at the transport layer, e.g., utilising a part of the sequence
number space that is encrypted or by verifying an encrypted token not
visible to an attacker. This would also mitigate against on-path
attacks. An additional processing cost can be incurred when decryption
is attempted before a receiver discards an injected packet.
The existence of open transport protocol standards and a research
and operations community with a history of independent observation and
evaluation of performance data encourage fairness and conformance to
those standards. This suggests careful consideration will be made over
where, and when, measurement samples are collected. An appropriate
balance between encrypting some or all of the transport header
information needs to be considered. Open data and accessibility to
tools that can help understand trends in application deployment, network
traffic, and usage patterns can all contribute to understanding security
challenges.
The security and privacy considerations in "A Framework for
Large-Scale Measurement of Broadband Performance (LMAP)" contain considerations for Active and Passive
measurement techniques and supporting material on measurement
context.
Addition of observable transport information to the path increases
the information available to an observer and may, when this information
can be linked to a node or user, reduce the privacy of the user. See the
security considerations of .
IANA Considerations
This document has no IANA actions.
Informative References
Measurement-based Protocol Design
European Conference on Networks and Communications, Oulu, Finland.
Reducing Internet Latency: A Survey of Techniques and Their
Merits
IEEE Communications Surveys & Tutorials, vol. 18, no. 3, pp. 2149-2196,
thirdquarter 2016
Bufferbloat: Dark Buffers in the Internet
Communications of the ACM, Vol. 55, no. 1, pp. 57-65
QUIC trace utilities
Commit 413c3a4
Revisiting the Privacy Implications of Two-Way Internet
Latency Data
Passive and Active Measurement
Acknowledgements
The authors would like to thank , , , , , ,
, , , , , ,
, , , , , , and members of TSVWG for their comments and
feedback.
This work has received funding from the European Union's
Horizon 2020 research and innovation programme under grant agreement No
688421 and the EU Stand ICT Call 4. The opinions expressed and
arguments employed reflect only the authors' views. The European
Commission is not responsible for any use that might be made of that
information.
This work has received funding from the UK Engineering and Physical
Sciences Research Council under grant EP/R04144X/1.