Conference PaperPDF Available

Slimfit — A HIP DEX compression layer for the IP-based Internet of Things


Abstract and Figures

The HIP Diet EXchange (DEX) is an end-to-end security protocol designed for constrained network environments in the IP-based Internet of Things (IoT). It is a variant of the IETF-standardized Host Identity Protocol (HIP) with a refined protocol design that targets performance improvements of the original HIP protocol. To stay compatible with existing protocol extensions, the HIP DEX specification thereby aims at preserving the general HIP architecture and protocol semantics. As a result, HIP DEX inherits the verbose HIP packet structure and currently does not consider the available potential to tailor the transmission overhead to constrained IoT environments. In this paper, we present Slimfit, a novel compression layer for HIP DEX. Most importantly, Slimfit i) preserves the HIP DEX security guarantees, ii) allows for stateless (de-)compression at the communication end-points or an on-path gateway, and iii) maintains the flexible packet structure of the original HIP protocol. Moreover, we show that Slimfit is also directly applicable to the original HIP protocol. Our evaluation results indicate a maximum compression ratio of 1.55 for Slimfit-compressed HIP DEX packets. Furthermore, Slimfit reduces HIP DEX packet fragmentation by 25 % and thus further decreases the transmission overhead for lossy network links. Finally, the compression of HIP DEX packets leads to a reduced processing time at the network layers below Slimfit. As a result, processing of Slimfit-compressed packets shows an overall performance gain at the HIP DEX peers.
Content may be subject to copyright.
Slimfit - A HIP DEX Compression Layer
for the IP-based Internet of Things
e Hummen, Jens Hiller, Martin Henze, Klaus Wehrle
Communication and Distributed Systems, RWTH Aachen University, Germany
Email: {lastname}
Abstract—The HIP Diet EXchange (DEX) is an end-to-end
security protocol designed for constrained network environments
in the IP-based Internet of Things (IoT). It is a variant of the
IETF-standardized Host Identity Protocol (HIP) with a refined
protocol design that targets performance improvements of the
original HIP protocol. To stay compatible with existing protocol
extensions, the HIP DEX specification thereby aims at preserving
the general HIP architecture and protocol semantics. As a result,
HIP DEX inherits the verbose HIP packet structure and currently
does not consider the available potential to tailor the transmission
overhead to constrained IoT environments. In this paper, we
present Slimfit, a novel compression layer for HIP DEX. Most im-
portantly, Slimfit i) preserves the HIP DEX security guarantees,
ii) allows for stateless (de-)compression at the communication
end-points or an on-path gateway, and iii) maintains the flexible
packet structure of the original HIP protocol. Moreover, we show
that Slimfit is also directly applicable to the original HIP protocol.
Our evaluation results indicate a maximum compression ratio
of 1.55 for Slimfit-compressed HIP DEX packets. Furthermore,
Slimfit reduces HIP DEX packet fragmentation by 25 % and thus
further decreases the transmission overhead for lossy network
links. Finally, the compression of HIP DEX packets leads to a
reduced processing time at the network layers below Slimfit. As a
result, processing of Slimfit-compressed packets shows an overall
performance gain at the HIP DEX peers.
KeywordsInternet of Things, Network Security, Key Manage-
ment, Compression, HIP, HIP DEX
Peer authentication and secure data transmission are vital
aspects for many application scenarios in the Internet of Things
(IoT) to prevent leakage of personal information or the exe-
cution of harmful actuation tasks, e.g., in building automation
or e-health systems. To secure end-to-end connections in the
IP-based IoT, a number of lightweight variants of existing
IP key management protocols have recently been proposed.
These protocol variants most notably include Datagram TLS
(DTLS) [1], the HIP Diet EXchange (DEX) [2], and minimal
IKEv2 [3]. With its four-way packet exchange, HIP DEX
thereby features a more concise handshake than DTLS with 6
round-trips and up to 15 packets and represents an alternative
to minimal IKEv2 for establishing an IPsec payload channel.
The specification of HIP DEX follows two central design
principles. On the one hand, the basic idea is to adapt the
original HIP protocol to the limited processing power, scarce
memory resources, and lossy network links in constrained
network environments. On the other hand, a key goal is to
maintain the general HIP architecture and protocol semantics
to remain compatible with the wide range of existing HIP
protocol extensions. These extension, among others, include
IoT Domain
IoT Domain
Local Network
Fig. 1. Resource-constrained devices (D) communicate with each other
and with local or Internet-based services (S) via a gateway (GW). Entities
belonging to an IoT domain are equipped with our Slimfit layer. Dashed arrows
indicate communication paths with compressed HIP DEX packets.
the support for host mobility and multi-homing [4] as well as
the integration of IPsec for transport security [5].
Although HIP DEX already provides a tailored retransmis-
sion mechanism and a refined session establishment handshake
that reduces the need for public-key-based security primitives
to a single Diffie-Hellman operation per peer, its packet
structure has not yet been subject to adaptations. As a result,
expendable information such as length information of static-
size parameters and per-parameter padding still needs to be
carried over constrained network links for each packet.
Our contributions in this paper are twofold. First, we
analyze the HIP DEX protocol with respect to its packet
content and identify to which extent this content can be omitted
or compressed before packet transmission. Second, we propose
Slimfit, a novel packet compression layer for HIP DEX. Slimfit
is located below the HIP DEX layer in the network stack
and affords stateless (de-)compression at an on-path gateway
and at the communication end-points (see Fig. 1). Moreover,
Slimfit preserves the flexible packet structure of the original
HIP protocol. Thus, it maintains compatibility with existing
and future protocol extensions. Finally, Slimfit is applicable to
the original HIP protocol without modifications.
This paper is structured as follows. Section II introduces
the network scenario and gives a brief overview of the HIP
DEX protocol. In Section III, we discuss and classify expend-
able protocol information that we identified during our HIP
DEX protocol analysis. We then present the design of our
proposed Slimfit compression layer and show how to compress
expendable HIP DEX protocol information in Section IV. In
Section V, we discuss the evaluation results of our proposed
Slimfit layer. We cover the security considerations when em-
ploying Slimfit in Section VI. Finally, Section VII explores
related work and Section VIII concludes our paper.
Initiator Responder
Fig. 2. Protocol diagram of the refined HIP DEX session establishment
handshake. Parameters marked bold can be compressed to a single bit. The
remaining parameters mainly contain random or cryptographic information.
We now briefly present the target network scenario for our
proposed Slimfit compression layer and give an overview of
the HIP DEX protocol as the basis of our work.
A. Network Scenario
As shown in Fig. 1, our target network scenario consists of
constrained devices in an IoT domain, services that are located
in a local network or the Internet, and gateways that intercon-
nect these network domains. We assume that communication
in the IoT domains takes place over constrained wireless links
as provided, e.g., by IEEE 802.15.4. Furthermore, we assume
that constrained devices are IP-enabled and are equipped with
6LoWPAN, an IPv6 adaptation layer for constrained network
environments that is standardized at the IETF [6]. The link
from the gateway to the local network or the Internet is a
commodity broadband connection.
With respect to the available resources, we assume con-
strained devices to be equipped with only a few MHz of
computational power, several kilobytes of RAM and several
tens of kilobytes of ROM. Moreover, these devices may be
battery-powered. Gateways and services, on the contrary, run
on common network and server hardware, respectively.
A key requirement for the design of HIP DEX [2] is to
maintain the general HIP architecture and protocol semantics.
Specifically, HIP DEX inherits the cryptographic HIP names-
pace that uses the public key of a device as its host identity
(HI). This namespace is used to build a new layer in the
network stack between the network and the transport layer.
The host identifier (HIT), a fixed-length representation of the
HI, serves as a stable device identifier at this layer.
To reduce protocol overhead, HIP DEX specifies a refined
session establishment handshake. This handshake consists of a
four-way packet exchange between an Initiator and a Respon-
der (see Fig. 2). The Initiator triggers the handshake with an
I1 message. The subsequent three messages then implement
a standard authenticated Diffie-Hellman (DH) key agreement.
HIP DEX thereby replaces the ephemeral DH keys and digital
signatures of the original HIP protocol with static DH keys for
mutual peer authentication and key agreement.
Additionally, HIP DEX specifies an aggressive retransmis-
sion mechanism to handle packet loss in constrained wireless
network environments. This mechanism requires the Initiator
to continually send I1 or I2 packets at short time intervals until
it receives the corresponding response packet.
01 2 345 6 78910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Next Header Header Length 0Packet Type Version Res. 1
Checksum Controls
Source HIT (128 bit)
Destination HIT (128 bit)
Type Length
further parameters...
Fig. 3. HIP packet structure consisting of a fixed header and a number
of signaling parameters. Each parameter is type-length-value encoded and
includes individual padding. Compressible information is marked in gray. The
fields marked light gray are handled at the 6LoWPAN layer.
The refined session establishment handshake and the ag-
gressive retransmission mechanism already cater to the limited
processing capabilities and the lossy link characteristics of
constrained network environments. However, the HIP DEX
specification still preserves the verbose packet structure of
the original HIP protocol and thus does not yet consider
unnecessary transmission overheads. As a result, already the
smallest HIP DEX handshake packet exceeds the maximum
payload size of an IEEE 802.15.4 frame and must be split into
multiple packet fragments for transmission. Packet fragmen-
tation in turn may be harmful [7] and leads to an increased
loss probability for the entire HIP DEX packet as the loss of
a single fragment results in the loss of the complete packet.
As a first step towards optimizing the transmission over-
head, we now analyze the HIP DEX packet structure and
classify the packet content by its space saving potential during
packet transmission. We thereby distinguish three types of
packet content: i) semantically irrelevant or static information
that can be omitted, ii) variable, redundant information that
can be represented in a compressed form, and iii) information
that cannot or should not be compressed.
Regarding the packet structure, HIP and HIP DEX share
a common packet wire-format for the signaling of protocol
information. As depicted in Fig. 3, each packet consists of
a fixed protocol header and a varying number of parameters.
These parameters carry the actual protocol information (see
Fig. 2). Both, the fixed header and the HIP parameters contain
information that often is useful in the scope of Internet-based
communication, but that constitutes unnecessary overhead dur-
ing packet transmission in constrained network environments.
A. Fixed HIP Header
Logically, the fixed HIP header is designed as an IPv6
extension header. As such, it starts with the mandatory Next
Header and Header Length fields and requires an 8 byte
alignment for the HIP DEX packet content. The HIP-specific
part of the header begins with information that is required
for the correct parsing of a packet, i.e., the packet type and
the protocol version. The two fixed bits enclosing these fields
are set for compatibility with other protocols, i.e., SHIM6.
The header further contains a checksum for early packet
verification and the HIP Controls field for information about
the packet structure and the host capabilities. As depicted in
Fig. 3, the fixed header ends with a HIT address pair that
identifies the source and destination at the HIP layer.
Compressibility: With 32bytes, the HIT pair requires the
majority of the 40 bytes for the fixed HIP header. This makes
HITs the primary target for compression. Regarding their struc-
ture, HITs have been designed to resemble IPv6 addresses. As
a result, HITs can, for example, be used in combination with
existing programming interfaces and as host referrals for IPv6-
enabled applications that are unaware of the underlying HIP
protocol layer [8]. HITs thereby consist of three components:
i) a fixed 28 bit prefix to distinguish HITs from standard IPv6
addresses, ii) a 4 bit HIT generation algorithm identifier, and
iii) a 96 bit HI representation that is derived with the indicated
generation algorithm. For HIP DEX, this generation algorithm
is specified as the left-truncation of the HI to 96 bit.
While highly useful to differentiate HITs and IPv6 ad-
dresses at the end-hosts, the HIT prefix constitutes static
information at the HIP layer and can therefore be omitted
from the HIP header before packet transmission. Likewise, HIP
DEX only supports a single HIT generation algorithm in its
protocol specification. This renders the generation algorithm
identifier omissible during transmission. Moreover, the HI
representation of a HIT can be derived if the packet contains
the HOST ID parameter carrying the corresponding HI. Thus,
the source HIT can be compressed as redundant information
in the R1 and I2 packets (see HOST ID parameters in Fig. 3).
Concerning the remaining header information, the manda-
tory IPv6 extension fields, i.e., the Next Header and Header
Length, can be modified with existing 6LoWPAN compression
facilities [9]. The HIP Controls field denotes a bit array with
each bit indicating distinct signaling information. Currently
only the highest order bit of the HIP Controls field is defined.
It indicates that the signaled HI is anonymous and should
therefore not be stored by the peer. As the overhead of multiple
anonymous identities likely is excessive for constrained de-
vices, the HIP Controls field typically does not contain any sig-
naling information and can therefore be omitted in constrained
network environments. Finally, the packet parsing information
should remain uncompressed in order to allow for immediate
identification of retransmitted packets and to afford checksum-
based integrity verification without prior decompression. To
this end, the checksum field should cover the compressed
packet content instead of the original HIP DEX packet.
B. HIP Parameters
The HIP parameters are type-length-value encoded and in-
clude separate padding information to guarantee 8 byte packet
alignment (see Fig. 3). Furthermore, parameters in a packet
follow a strict numeric order according to their type number.
Hence, the position of a parameter in a packet is well-defined.
A gateway or the communication partner can use this
encoding of signaling information to parse and process packet
content based on parameter type numbers. Moreover, they
can skip over unsupported or unneeded signaling information
based on the length information of the respective parameters.
The HIP DEX specification builds on this packet structure by
defining a limited set of mandatory parameters that must be
understood by the peer in order to set up a security association.
Additional protocol extensions can then easily be specified by
adding new parameters that are optional to be supported.
Compressibility: Although required for the alignment of
the IPv6 extension header, the semantically irrelevant per-
parameter padding unnecessarily adds to the overall packet
size. Hence, this padding should be omitted during packet
transmission. Moreover, the length field of static-size manda-
tory parameters contains redundant information as the length
of these parameters can be derived from the protocol spec-
ification. Likewise, the type field of mandatory parameters
constitutes redundant information for packets without optional
parameters as the order of mandatory parameters is well-
defined. Hence, while necessary for a consistent parameter
layout, these redundancies render the type and length fields of
mandatory parameters omissible during packet transmission.
The omission of the type and length fields for optional param-
eters is not possible as their occurrence in a packet may vary
for the same packet type. Furthermore, optional parameters
are not necessarily supported by the communication partner,
rendering length information unavoidable for packet parsing.
Regarding compressible parameter content, HIP and HIP
DEX both provide cipher suite negotiation mechanisms for
the peers to agree on a mutually supported cipher suite and to
afford modular protocol evolvability. However, especially in
the context of constrained devices, the set of supported cipher
suites typically is limited to a small number of cryptographic
primitives and cipher modes to reduce ROM requirements.
This allows to compress default values in the negotiation
parameters before packet transmission. Still, this compression
should not only be applicable to the status quo, but should
evolve with future security recommendations.
As shown in the previous section, the HIP DEX packet
structure still includes a substantial amount of unnecessary
overhead during packet transmission. Hence, we propose Slim-
fit, a novel compression layer that tailors HIP DEX transmis-
sion overhead to constrained network environments. Notably,
Slimfit is only executed at network entities that belong to the
constrained IoT domain, i.e., constrained communication end-
points or gateways (see Fig. 1). Thus, end-points and network
elements such as firewalls that are located outside a constrained
network domain remain oblivious to our Slimfit layer.
We now show how Slimfit integrates in the processing of
HIP DEX signaling packets. We then present the compression
mechanisms that Slimfit employs in order to reduce the over-
head of the fixed HIP header and the HIP signaling parameters.
A. Integration of Slimfit in the Network Stack
As illustrated in Fig. 4, our proposed Slimfit compression
layer is located between the network and the HIP DEX layer
in the network stack. Furthermore, Slimfit does not directly
interact with the HIP DEX layer. This design enables a Slimfit-
agnostic HIP DEX protocol implementation and a transparent
integration of Slimfit in the HIP DEX packet processing.
When the HIP DEX layer at a sending node inside an IoT
domain generates a new packet, this packet must pass through
our Slimfit compression layer before transmission towards the
IoT domain Internet
H M1 O1 M2
H O1
H M1 O1 M2
H O1
H M1 O1 M2
IP H M1 O1 M2
6 IP H O1
Fig. 4. Integration of our proposed Slimfit compression layer in the network
stack. Arrows indicate the processing flow for a HIP DEX packet with an
on-path gateway. Slimfit compresses the fixed HIP header (H), reorders the
mandatory (Mi) and optional (Oj) parameters, and compresses or omits these
parameters. The dashed arrow indicates that the HIP DEX packet skips the
Slimfit layer and is forwarded without compression at the 6LoWPAN layer.
HIP DEX peer. For a constrained device, this can be achieved
by placing our compression layer at the corresponding code
point in the embedded network stack. A similar integration
can be achieved for commodity operating systems by hooking
into the network stack via dedicated networking facilities such
as netfilter for Linux. Slimfit then performs the required HIP
DEX packet compression and informs the peer that the packet
structure has been modified. To this end, it indicates the used
compression mechanisms in the unoccupied bits of the HIP
Controls field (see Fig. 5). The Slimfit layer then passes the
compressed packet to the network layer for further processing.
As shown in Fig. 4, the compressed HIP DEX packet
then reaches the 6LoWPAN layer. Here, the HIP DEX-related
IPv6 extension header is further compressed as defined in [9].
Specifically, 6LoWPAN recomputes the Header Length as
a multitude of 1 byte units. Thus, compressed HIP DEX
packets no longer require padding to obey the 8 byte extension
header alignment. To indicate this next header compression, the
6LoWPAN layer prepends an additional byte to the HIP DEX
header (i.e., the LOWPAN NHC field [9]). We propose to use
one of the free Extension Header IDs (EID) in this header,
e.g., an EID value of 6, to indicate the payload-independent
6LoWPAN extension header compression described above.
After 6LoWPAN processing has finished, the packet traverses
the network stack without further modification.
On-path nodes, that merely forward packets towards their
final destination, do not process HIP DEX packet content.
Hence, no special attention is required for compressed packets
on the forwarding path. This is unless a gateway is reached that
bridges the constrained network domain with an unconstrained
network such as the Internet. In this case, Slimfit decompresses
the packet after it has been processed at the network layer of
the gateway as depicted in Fig. 4. To this end, our proposed
compression layer directly hooks into the network layer, e.g.,
using netfilter. Likewise, if a constrained peer receives a
compressed HIP DEX packet, Slimfit decompresses this packet
before passing it to the HIP DEX layer. Slimfit thereby first
determines the indicated compression mechanisms and then
applies their respective decompression counterparts.
B. Compression of the Fixed HIP Header
As depicted in Fig. 5, Slimfit uses the 2 byte HIP Controls
field to signal the compression mechanisms that were applied
to the original HIP DEX packet. Still, the second byte of this
HIP Controls
HIT compr.
HIT algo.
Param. compr.
DH Group
HIP Cipher
HIT Suite
TP Format
Neg. Profile
01 2 345 6 78910 11 12 13 14 15
Fig. 5. The compression bits set in the HIP Controls field indicate a
HIP DEX packet with maximum Slimfit compression. Negotiation parameter
compression flags are marked with an asterisk. Omitted content is highlighted
gray. Bits 11 to 14 remain unused.
field can be omitted before transmission if all flags at this byte
are unset. Slimfit signifies this HIP Controls compression in
the first bit of the compressed HIP Controls field. Moreover,
Slimfit removes the static prefix of the source and destination
HIT for all HIP DEX packets and omits the HI representation
of the source HIT for R1 and I2 packets. This compression is
indicated by setting the second bit in the HIP Controls field to
1. Finally, Slimfit elides the HIT generation algorithm for the
HIP DEX algorithm “ECDH/DEX” and indicates this omission
in bit 3. We defer the discussion about the evolvability of the
latter compression step to Section IV-D.
C. HIP Parameter Compression
To achieve a high compression rate for the general HIP
parameter fields (i.e., type, length, and padding), Slimfit first
reorders the parameters of the uncompressed HIP DEX packet.
More precisely, Slimfit breaks the strict parameter order of HIP
DEX and moves the optional parameters behind the mandatory
ones (see Fig. 4). Mandatory parameters thereby maintain their
relative position towards each other. Hence, their order remains
well-defined for each packet type. The order among the op-
tional parameters is irrelevant for the subsequent compression
steps. Slimfit then removes the padding information from all
parameters. Furthermore, Slimfit dismisses the length field
from all fixed-length mandatory parameters and removes the
type field from all mandatory parameters. The latter omission
is only possible due to the new parameter structure in Slimfit-
compressed HIP DEX packets. Slimfit indicates this general
parameter compression by setting the fourth HIP Controls
bit to 1. The compression is unambiguously reversible by
performing the above steps in reverse order.
Slimfit also represents entire HIP parameters as flags in the
HIP Controls field (see omitted parameter M1in Fig. 4 and
asterisks in Fig. 5). Specifically, Slimfit can compress the four
negotiation parameters DH GROUP LIST, HIP CIPHER,
single bit each if the respective parameter only contains default
values. We now briefly discuss the main reasons behind our
choice of the specific default values for these parameters and
refer to Section IV-D for a description of a simple mechanism
that provides evolvability of selected default values.
The DH GROUP LIST parameter contains a list of well-
known DH groups that a device supports. With HIP DEX,
this parameter is limited to elliptic curve-based DH groups.
Additionally, the possible values of this parameter can further
be narrowed down when considering recent recommendations
by NIST [10]. Specifically, NIST P-256 is the smallest elliptic
curve that is recommended for use after 2013 and that is also
defined for HIP DEX. This curve furthermore is expected to
be secure until 2030 according to NIST. Hence, we expect
constrained devices to use this curve in the HIP DEX hand-
shake and compress the DH GROUP LIST parameter if it
consists of the corresponding DH group ID. For each of the
remaining negotiation parameters, HIP DEX currently only
defines a single valid suite ID. Hence, we propose to declare
these IDs as default values for the HIP DEX protocol and
compress these parameters as individual HIP Controls flags.
D. Compression Evolvability
On the one hand, Slimfit packet compression reduces HIP
DEX packet sizes by omitting and compressing general aspects
of the HIP packet structure. This type of compression remains
applicable as long as the packet structure does not fundamen-
tally change in future protocol iterations. On the other hand,
Slimfit compresses header fields and parameters that contain
default values. This choice of default values will not hold
indefinitely and may also be scenario-specific. To enable the
modification of default values, we additionally propose the
use of profiles that define a new set of default values and
overwrite our specific choice above. We reserve three bits in
the HIP Controls field to indicate the use of such profiles by
means of profile IDs. As Slimfit is only deployed inside IoT
domains, profiles and their IDs may be domain-specific. Most
importantly, our negotiation parameter compression can evolve
over time by leveraging these profiles.
E. Applicability of Slimfit to the original HIP protocol
HIP and HIP DEX share the same packet structure and
define mandatory as well as optional parameters for the spec-
ified protocol exchanges. Hence, our proposed fixed header
and general parameter compression also apply to the original
HIP protocol. However, both protocols differ with respect to
their target network scenarios. While HIP DEX focuses on
constrained network environments, the original HIP protocol
aims at securing communication between comparably powerful
Internet hosts. As a result, the cryptographic primitives em-
ployed in HIP and, thus, the content of the negotiation param-
eters differ from the default values for HIP DEX. To achieve
similar compression results for the original HIP protocol as for
HIP DEX, we propose to define a new negotiation parameter
profile as discussed above. Notably, this profile must capture
the potential use of RSA or DSA for the HIs of the peers as
the main difference between HIP and HIP DEX.
For our evaluation, we implemented the HIP DEX protocol
as specified in [2] for the Contiki OS and extended this
implementation with our Slimfit compression layer. The HIP
DEX implementation employs the relic1library with elliptic
curve NIST P-256 for public-key operations. We used Zolertia
Z1 motes that are equipped with a 16 MHz MSP430 microcon-
troller, 8 kB of RAM, 92 kB of ROM, and an IEEE 802.15.4
radio interface as our evaluation platform for constrained
devices. Moreover, we implemented a simple Linux application
that utilizes the generic compression libraries lz77,lzma,lzss,
and zlib (which implements deflate) to compress real HIP DEX
packet traces. We used this application as a means to compare
payload size (byte)
number of fragments
Fig. 6. Packet sizes of the HIP DEX handshake and the session tear down
exchange (marked with a grey background) for the standard protocol, with zlib
compression, and with Slimfit compression. Dashed lines indicate the number
of fragments required for the transmission of the HIP DEX packet content.
the compression ratio of our protocol-specific Slimfit layer
against alternative generic compression approaches.
Regarding packet transmission, we assumed that IoT net-
works employ link layer security to prevent network attacks
against, e.g., the local routing structure. To simulate the
corresponding header overhead, we decreased the payload size
at the 6LoWPAN adaptation layer by 21 bytes. As a result,
our evaluation shows packet fragmentation that would occur in
secure network scenarios. In such scenarios, the first fragment
of a packet contains up to 32 bytes and subsequent fragments
at maximum 72 bytes of HIP DEX packet content.
A. Transmission Reductions
To quantify the transmission reductions of our Slimfit
compression layer, we measured the packet sizes of a standard
HIP DEX handshake and a subsequent session tear down be-
tween two wirelessly connected Z1 motes. We then compared
this baseline against the size of Slimfit-compressed packets
in order to determine the compression ratio of our Slimfit
layer. Furthermore, we compared the Slimfit compression ratio
to the compression ratio of generic compression algorithms
when applied to HIP DEX packets. To this end, we ran
our Linux application with the above algorithms against a
captured packet trace consisting of 100 HIP DEX handshake
and session tear down exchanges. For all generic algorithms
we used the highest available level of compression. During our
evaluation, we found that, regarding the considered algorithms,
zlib achieves the best overall space savings for HIP DEX
packets. Hence, we focus our discussion on this algorithm.
As shown in Fig. 6, our proposed Slimfit layer is able
to compress all HIP DEX packets and outperforms the zlib
algorithm. More precisely, Slimfit achieves a compression ratio
that ranges from 1.36 for I2 packets up to 1.55 for I1 packets.
Note that these compression ratios can be achieved for any HIP
DEX packet containing the previously defined default values.
In contrast, zlib has a considerably lower compression ratio of
1.18 in its best case and even adds a small overhead of up to
7 bytes to short HIP DEX packets (see I2 and R2 in Fig. 6).
This is due to the constant 11 byte overhead introduced by
zlib. Moreover, we observed a maximum standard deviation
of 0.69 byte for zlib. This signifies the modest dependency of
zlib on the actual packet content.
Regarding the overall transmission overhead, our Slimfit
compression layer decreases the HIP DEX handshake size
0 10 20 30 40 50 60 70 80
loss ratio (%)
transmitted packet size (byte)
Fig. 7. Overall handshake transmission overhead for the standard HIP DEX
protocol and with our Slimfit compression layer for different packet loss
probabilities. Note the logarithmic scale of the y-axis. Error bars depict the
standard deviation of our measurements.
from 536 bytes to 372 bytes. This is a reduction of 30.6 %. In
contrast, zlib only achieves a reduction of 7.14 % on average.
For the HIP DEX session tear down, Slimfit decreases the over-
head from 160 bytes to 114 bytes, a reduction of 28.75%. With
zlib, instead, the transmission overhead increases by a total of
6.9 bytes (4.31 %). Thus, Slimfit substantially outperforms the
considered generic compression mechanisms.
Impact on packet fragmentation. Our proposed Slimfit layer
decreases HIP DEX handshake fragmentation by 3fragments
(see I1, R1, and I2 packets in Fig. 6). This constitutes a reduc-
tion of 25 % compared to the uncompressed handshake. As a
result, the radio utilization and the packet processing overhead
caused by HIP DEX on the forwarding path is reduced due to
the transmission of fewer packet fragments. Additionally, the
reduced packet fragmentation has a positive impact on HIP
DEX packet retransmissions. To further quantify this aspect,
we analyzed the overall size of a HIP DEX handshake includ-
ing retransmitted packets between an Initiator and a Responder
in the Cooja network simulator for Contiki. We thereby placed
both nodes within the direct radio range of each other and
measured the average transmission overhead for end-to-end
loss probabilities ranging from 0 % to 80 % for individual
packet fragments. For each loss probability we performed 5000
handshakes. We decided for simulation over a real testbed
to achieve well-defined packet loss probabilities without side-
effects on the wireless medium. Note that already the loss of a
single fragment results in the loss of the entire corresponding
HIP DEX packet as fragment transmission follows the best
effort semantics of IPv6.
As depicted in Fig. 7, the decreased packet fragmentation
with Slimfit considerably reduces HIP DEX retransmissions.
The main reason is that the overall loss probability of a
compressed HIP DEX packet decreases as the packet consists
of fewer fragments compared to an uncompressed packet. As a
result, the average transmission overhead with Slimfit already
constitutes about 63.5 % of an uncompressed handshake at a
loss probability of 20 %. Notably, this ratio further improves
for increasing loss probabilities. Hence, Slimfit not only re-
duces the packet size of individual HIP DEX packets, but also
improves the HIP DEX performance for lossy network links.
B. Processing Overhead
To evaluate the computation overhead introduced by our
proposed Slimfit layer, we measured the packet processing
I (-,1) R (1,2) I (2,3) R (3,4) I (4,-) I (-,5) R (5,6) I (6,-)
processing time (ms)
Lower layers
Fig. 8. Processing time of the HIP DEX handshake and the session tear down
for the standard HIP DEX protocol (H) and with our Slimfit compression layer
(S) for the Initiator (I) and the Responder (R). Numbers in brackets denote
i-th packet processing and (i+ 1)-th packet generation.
time on two wirelessly connected Z1 motes over 100 measure-
ment runs. More precisely, we considered the processing time
for packets belonging to the HIP DEX handshake and to the
session tear down. We thereby measured the computation time
at the HIP DEX layer, our Slimfit layer, and at the layers below
Slimfit in the Contiki network stack. Concerning the lower
layers, we restricted our evaluation to the processing overhead
of outbound HIP DEX packets to prevent overhead misattribu-
tion for inbound packet fragments that belong, e.g., to the RPL
routing protocol. Hence, inbound HIP DEX packets commonly
involve additional computation overhead at the lower layers
compared to the overhead depicted in Fig. 8. The standard
deviation of our measurements was below 23.09ms (1.25 %)
for public-key-based operations of 1840.86 ms on average and
below 0.09 ms for the remaining measured operations.
Fig. 8 illustrates the average packet processing time for
uncompressed and Slimfit-compressed HIP DEX packets. Cu-
riously, Slimfit packet compression does not result in an overall
performance penalty. Instead, it even leads to a modest overall
performance gain that amounts to 6.58 ms for the complete
HIP DEX handshake and to 0.83 ms for the session tear
down exchange. During our analysis of this phenomenon, we
found that this performance gain mainly stems from a reduced
computation overhead at the lower layers that outweighs the
low processing overhead at our Slimfit layer (see “Lower lay-
ers” and “Slimfit” in Fig. 8). Specifically, Slimfit-compressed
packets require less computation overhead at the 6LoWPAN
layer due to reduced packet fragmentation. Likewise, the MAC
layer has to handle less IEEE 802.15.4 frames for tasks such as
transmission scheduling and carrier sensing. The correspond-
ing performance gains amount to 8.61 ms for the complete
HIP DEX handshake and to 1.67 ms for the session tear down.
These gains would further increase when also considering the
overhead of inbound HIP DEX packets at the lower layers.
At the same time, the additional processing time at our
Slimfit layer only constitutes a minimum of 0.12 ms for the
compression of an I1 packet and a maximum of 0.73 ms for
the decompression of an R1 packet as well as the compression
of an I2 packet (see “I (-,1)” and “I (2,3)” in Fig. 8).
Hence, the performance benefits at the lower layers outweigh
the computation cost at our Slimfit layer, resulting in an
overall performance gain for Slimfit-compressed HIP DEX
packets. To conclude, our Slimfit compression layer not only
decreases transmission overhead and packet fragmentation, but
also decreases overall processing time of HIP DEX packets.
Extension ROM (byte) RAM (byte)
Contiki OS incl. HIP DEX 58659 7624
+ Slimfit compression layer 61157 (+2498)7624 (+0)
C. RAM and ROM Overhead
To derive RAM and ROM estimates for our proposed
Slimfit layer, we first analyzed the Contiki binary of the
unmodified HIP DEX protocol with the msp430-size tool. We
then compared the results to the binary that also includes our
Slimfit layer. Table I summarizes the results of our analysis.
Most importantly, the unchanged, static RAM overhead
proves that Slimfit indeed is stateless. This claim is further
substantiated by the fact that our implementation does not
use dynamic memory allocation, e.g., by means of malloc,
to maintain per-connection state. However, our Slimfit layer
generates a ROM overhead of about 2.5kB. This constitutes a
modest and, in fact, the only tradeoff of our proposed Slimfit
compression layer that we identified during our evaluation.
We now briefly discuss attacks that an adversary can mount
against devices that employ our Slimfit compression layer.
Exploiting the Slimfit-compressed packet structure. Con-
tent compression before encryption has been shown to leak
information about the uncompressed content [11]. Slimfit only
compresses or omits packet content that would otherwise be
transmitted in plaintext. Hence, it is immune against such side-
channel attacks and does not reveal sensitive packet content.
Likewise, packet loss and out-of-order packets may cause
an invalidation of the compression context for stateful com-
pression mechanisms [12]. An adversary could aim at ex-
ploiting this fact to prevent a Slimfit-compressed HIP DEX
packet exchange from completing successfully. However, Slim-
fit packet compression does not require compression state
across HIP DEX packets and thus is resistant to such attacks.
Targeting the Slimfit packet processing overhead. Any
networked adversary could flood a target device with Slimfit-
compressed I1 or I2 packets in a Denial of Service (DoS)
attack targeting the resulting packet processing overhead. As
shown in Sec. V-B, Slimfit modestly reduces the overall packet
processing overhead and, thus, would in fact decrease the
impact of this attack. Hence, Slimfit does not open a new
vector of attack compared to the standard HIP DEX protocol.
Impact of Slimfit on the DoS protection in HIP DEX.
The HIP DEX protocol offers two handshake mechanisms
to protect the Responder against a DoS attack targeting the
protocol computation and memory overhead. On the one hand,
HIP DEX enables the Responder to delay state creation until
the reception of the I2 packet. As a result, the Responder
only has to store HIP DEX session state after a successful
authentication of the Initiator. On the other hand, HIP DEX
employs a puzzle mechanism that enables the Responder to
demand an adjustable resource commitment from the Initiator.
This allows the Responder to only invest resources into the
processing of a received I2 packet after the Initiator has
committed to the handshake by solving the issued puzzle.
Slimfit does not alter these DoS protection properties of the
HIP DEX protocol. Specifically, as Slimfit does not require any
compression context to be maintained across packets, a Slimfit-
enabled Responder can remain stateless before receiving an
I2 packet. Moreover, Slimfit does not modify the puzzle
mechanism and even decreases the processing cost of an I2
packet at the network layers that are located below Slimfit.
Hence, Slimfit-enabled peers are equally protected against DoS
attacks as peers that run the uncompressed HIP DEX protocol.
Slimfit and on-path network entities. Slimfit enables gate-
ways that are located at the edge of the IoT domain to perform
HIP DEX packet compression. An adversary could flood these
gateways with Slimfit-compressed HIP DEX packets in order
to exhaust the available gateway resources. As a result, the
adversary could hamper packet forwarding between the IoT
domain and the external network. However, such flooding
attacks would be limited to the computation overhead of
Slimfit as it does not require state information for packet
compression. Yet, the modest processing cost of Slimfit renders
the 6LoWPAN layer a more attractive target.
On-path security appliances such as firewalls may aim at
inspecting HIP DEX packets to provide additional protection
for the IoT domain. Slimfit changes the packet structure of
HIP DEX packets for packet compression. Hence, security
appliances that support the filtering of HIP DEX, do not
support Slimfit-compressed packets per se. However, Slimfit
compression is limited to the IoT domain and, thus, does
not impact middleboxes that are located in external networks.
Furthermore, security appliances inside the IoT domain can
still identify the packet length, type, and destination of a
Slimfit-compressed HIP DEX packet without decompression.
Hence, blocking of flooding attacks targeting a specific peer
or dropping of exceedingly large packets can still be achieved.
However, if filtering should occur on additional packet infor-
mation, the packet must first be decompressed. Yet, the over-
head of this operation is modest as shown in our evaluation.
For our discussion of related work, we distinguish three
research directions: i) protocol-specific compression mecha-
nisms for the IP-based IoT, ii) generic protocol compression
schemes, and iii) further related protocol mechanisms.
Several protocol-specific compression mechanisms have
recently been proposed in the context of the IoT. The compres-
sion of IPv6 headers and extension headers as well as of UDP
headers with 6LoWPAN is standardized in [6], [9]. Regarding
the compression of IP security protocols, Raza et al. pre-
sented initial ideas on 6LoWPAN-compressed DTLS in [13].
Likewise, header compression for IPsec payload channels was
presented in [14], [15] and was recently proposed for stan-
dardization in [16]. Our work distinguishes itself from these
approaches by introducing an evolvable compression scheme
for default values and by achieving high compression ratios via
design-level changes of the packet structure for compression
purposes. Moreover, IPsec compression is complementary to
our work and improves the applicability of IPsec as the default
payload protection mechanism for HIP DEX.
Generic protocol compression schemes have been pro-
posed for IPv6 as well as for upper layer protocols. The
IP Payload Compression Protocol (IPComp) [17] allows to
apply generic compression algorithms, including deflate, to
IPv4 and IPv6 packets in order to compress their payload.
As shown in our evaluation, such generic algorithms do not
leverage domain knowledge about the packet structure and
thus typically achieve lower compression ratios for HIP DEX
packets than our Slimfit layer. The RObust Header Com-
pression (ROHC) framework [18] consists of protocol-specific
compression profiles that, in contrast to our work, often employ
stateful compression approaches. Slimfit could be integrated
into ROHC as a stateless compression profile for HIP and HIP
DEX. Recently, the generic 6LoWPAN-GHC header compres-
sion mechanism has been proposed for standardization [19].
This approach introduces a generic compression algorithm that
facilitates protocol-specific dictionaries. However, while apply-
ing domain knowledge similar to our approach, 6LoWPAN-
GHC is limited to compressing the original HIP DEX packet
structure and cannot modify the packet structure to achieve a
higher compression efficiency. Still, such a generic approach
could be used in combination with Slimfit to further compress
optional parameters that, e.g., contain certificates.
Apart from packet compression, further protocol mecha-
nisms have been proposed that aim at reducing protocol trans-
missions. In [20], [21], the authors propose TLS extensions that
allow clients to cache static server information such as pub-
lic keys. Furthermore, session resumption mechanisms [20],
[22], [23] allow to decrease the transmission overhead of
the protocol handshake. However, in contrast to our stateless
approach with Slimfit, these mechanisms require an initial
handshake to establish the necessary state information at the
communication end-points. Still, these mechanisms can be
used in combination with our Slimfit layer to further reduce
the HIP DEX transmission overhead after an initial handshake.
The use of standard protocols in the IoT affords interop-
erable communication between constrained IoT devices and
services. However, device and network constraints necessitate
the employed protocols to be tailored towards the special
requirements of IoT network environments. In this paper2, we
analyzed and reduced the transmission overhead of HIP DEX,
a key management protocol that aims at securing end-to-end
connections in the IoT. In our protocol analysis, we identified
expendable information in the HIP DEX packet structure.
While useful in the scope of Internet-based communication, the
transmission of this information is undesirable in constrained
network environments. Our proposed Slimfit compression layer
removes the identified redundancies in the HIP DEX packet
before transmission by: i) applying domain knowledge derived
from the protocol specification, ii) modifying the packet struc-
ture to increase compression efficiency, and iii) introducing
an evolvable compression scheme for cipher suite negotiation
parameters. Our evaluation shows that Slimfit achieves a HIP
DEX packet compression ratio of up to 1.55 and reduces
packet fragmentation by 25 %. Moreover, Slimfit considerably
decreases HIP DEX retransmissions and modestly reduces
2This research is funded by the DFG Cluster of Excellence on Ultra High-
Speed Mobile Information and Communication (UMIC).
the overall packet processing overhead. Notably, the marginal
ROM overhead for our Slimfit implementation is the only
tradeoff for the achieved transmission and computation gains.
Finally, our design is not limited to the packet compression
of the status quo, but additionally considers the adaption of
Slimfit to future security recommendations. To conclude, the
integration of our Slimfit layer in the network stack is highly
beneficial when securing the IoT with HIP or HIP DEX.
[1] E. Rescorla and N. Modadugu, “Datagram Transport Layer Security
Version 1.2,” RFC 6347, IETF, IETF, 2012.
[2] R. Moskowitz, “HIP Diet EXchange (DEX),” draft-moskowitz-hip-dex-
00 (WiP), IETF, 2012.
[3] T. Kivinen, “Minimal IKEv2,” draft-kivinen-ipsecme-ikev2-minimal-01
(WiP), IETF, 2012.
[4] P. Nikander, T. Henderson, C. Vogt, and J. Arkko, “End-Host Mobility
and Multihoming with the Host Identity Protocol,” RFC 5206, IETF,
[5] P. Jokela, R. Moskowitz, and P. Nikander, “Using the Encapsulating Se-
curity Payload (ESP) Transport Format with the Host Identity Protocol
(HIP),” RFC 5202, IETF, 2008.
[6] G. Montenegro, N. Kushalnagar, J. Hui, and D. Culler, “Transmission
of IPv6 Packets over IEEE 802.15.4 Networks,” RFC 4944, IETF, 2007.
[7] R. Hummen, J. Hiller, H. Wirtz, M. Henze, H. Shafagh, and K. Wehrle,
“6LoWPAN Fragmentation Attacks and Mitigation Mechanisms,” in
Proc. of ACM WiSec, 2013.
[8] T. Henderson, P. Nikander, and M. Komu, “Using the Host Identity
Protocol with Legacy Applications,” RFC 5338, IETF, 2008.
[9] J. Hui and P. Thubert, “Compression Format for IPv6 Datagrams over
IEEE 802.15.4-Based Networks,” RFC 6282, IETF, 2011.
[10] E. Barker, W. Barker, W. Burr, W. Polk, and M. Smid, “Recommenda-
tion for Key Management – Part 1: General (Revision 3),” NIST Special
Publication 800-57, NIST, 2012.
[11] J. Kelsey, “Compression and information leakage of plaintext,” in Fast
Software Encryption, ser. LNCS. Springer Berlin Heidelberg, 2002.
[12] M. Degermark, H. Hannu, L. Jonsson, and K. Svanbro, “Evaluation of
CRTP Performance over Cellular Radio Links,IEEE Pers. Commun.,
vol. 7, no. 4, 2000.
[13] S. Raza, D. Trabalza, and T. Voigt, “6LoWPAN Compressed DTLS for
CoAP,” in Proc. of IEEE DCOSS, 2012.
[14] J. Granjal, E. Monteiro, and J. Sa Silva, “Enabling Network-Layer Secu-
rity on IPv6 Wireless Sensor Networks,” in Proc. of IEEE GLOBECOM,
[15] S. Raza, S. Duquennoy, T. Chung, D. Yazar, T. Voigt, and U. Roedig,
“Securing Communication in 6LoWPAN with Compressed IPsec,” in
Proc. of IEEE DCOSS, 2011.
[16] S. Raza, S. Duquennoy, and G. Selander, “Compression of IPsec AH
and ESP Headers for Constrained Environments,” draft-raza-6lowpan-
ipsec-00 (WiP), IETF, 2013.
[17] A. Shacham, B. Monsour, R. Pereira, and M. Thomas, “IP Payload
Compression Protocol (IPComp),” RFC 3173, 2001.
[18] K. Sandlund, G. Pelletier, and L.-E. Jonsson, “The RObust Header
Compression (ROHC) Framework,” RFC 5795, IETF, 2010.
[19] C. Bormann, “6LoWPAN Generic Compression of Headers and Header-
like Payloads, draft-bormann-6lowpan-ghc-06 (WiP), IETF, 2013.
[20] H. Shacham, D. Boneh, and E. Rescorla, “Client-side caching for TLS,”
[21] S. Santesson and H. Tschofenig, “Transport Layer Security (TLS)
Cached Information Extension,” draft-ietf-tls-cached-info-14 (WiP),
IETF, 2013.
[22] R. Hummen, J. H. Ziegeldorf, H. Shafagh, S. Raza, and K. Wehrle, “To-
wards Viable Certificate-based Authentication for the Web of Things,”
in Proc. of ACM HotWiSec, 2013.
[23] R. Hummen, H. Wirtz, J. H. Ziegeldorf, J. Hiller, and K. Wehrle,
“Tailoring End-to-End IP Security Protocols to the Internet of Things,
in Proc. of IEEE ICNP, 2013.
... The Host Identity Protocol (HIP) [15][16][17][18][19][20] can a suitable solution for IoT devices considering the security and privacy requirements of stationary and mobile IoT systems. The HIP can uniquely identify devices both in the mobile and stationary IoT environments and can be utilized to ensure identity and communication security. ...
... The authors in [19] propose a compression layer (Slimfit) in the protocol stack to reduce the size of the HIP headers. The Slimfit layer is introduced between the HIP and network layer. ...
... In [19], HIP headers are compressed to reduce the communication overheads for sending and receiving messages. However, there is an increase in the energy consumption and runtime both on the sender and receiver devices as every outgoing packet is compressed by a sender and every incoming packet is decompressed by a receiver. ...
Full-text available
The Host Identity Protocol (HIP) has emerged as the most suitable solution to uniquely identify smart devices in the mobile and distributed Internet of Things (IoT) systems, such as smart cities, homes, cars, and healthcare. The HIP provides authentication methods that enable secure communications between HIP peers. However, the authentication methods provided by the HIP cannot be adopted by the IoT devices with limited processing power because of the computation-intensive cryptographic operations involved in hash generation, signature validation, and session key establishment. Moreover, IoT devices cannot utilize the HIP as is to communicate securely in the low power and lossy networks as there is a considerable communication overhead, such as packet fragmentation and reassembly, for exchanging certificates over a lossy link. Additionally, the use of static host identifiers makes IoT devices vulnerable to cyber espionage and user-targeted attacks. In this article, we propose an authentication scheme, P-HIP, that protects the identity privacy of an IoT device by enabling the device to compute and use unique host identifiers from networks to networks and sessions to sessions. To make the HIP suitable for resource-constrained IoT devices, P-HIP provides methods that unburden IoT devices from computation-intensive operations, such as modular exponentiation, involved in authentication and session-key exchange. Additionally, P-HIP minimizes the communication overheads for exchanging certificates in lossy networks. We implement a prototype of P-HIP on Contiki enabled IoT that shows P-HIP can reduce computation costs, communication overheads, and the session-key establishment time when used by low-powered devices in a lossy network.
... Even if HIP (both HIP BEX and Diet-HIP) is based on a few signaling messages to establish a secure association between two peers on the Internet, these messages are very long and the expensive cost in terms of energy consumption cannot be ignored and also the fragmentation/reassembly rate resulting from the communication of similar messages within a 6LoWPAN network. To resolve this problem, the authors in [17] propose a compressed layer for HIP DEX named Slimfit. The compression concerns some fields in the HIP header and signaling parameters for the reduction of the communication cost. ...
... In this section, we discuss and compare previous HIP solutions with LC-DEX. Slimfit [17] cannot be suitable for lossy IoT networks because it does not deal with the minimization of the communication overhead that we consider an obstacle on the real-time application on the IoT. In contrast, LC-DEX is considering this issue and shows better results in communication energy cost compared with previous works. ...
Full-text available
The high level of security requirements and low capabilities of constrained devices that are connected to the Internet of Things (IoT) constitute a new challenge in terms of proposing an authentication solution that deals with the problem of energy constraints. The Host Identity Protocol Diet EXchange (HIP DEX) is primarily designed to be suitable for constrained devices and designed to be resistant to Denial of Service (DoS) and man-in-the-middle (MITM) attacks. In this paper, we propose an efficient saving energy solution to secure end-to-end (E2E) communications based on the compression of the IPv6 over Low Power Wireless Personal Area Networks (6LoWPAN) header for HIP DEX packets. We implement our solution in an IoT based-WSN over Constrained Application Protocol (CoAP) in the application layer and Routing Protocol for Low power and lossy networks (RPL) in the routing layer. We also propose a novel distribution model that minimizes the number of signaling messages. Both proposed compression and distribution models for HIP DEX combined with an original implementation of an opportunistic association establishment of the handshake, constitute an efficient security solution for IoT. We called our solution Lightweight Compressed HIP DEX in the IoT (LC-DEX).
... A secure multicast architecture for CoAP is proposed by Park et al. [49]. The study presented a design to secure both group communication and pairwise communication [41] Lithe et al. [42] Hummen et al. [43] ROHC [44] Bormann et al. [45] Granja (2010) [46] Granja (2015) [47] Carrillo et at. [48] Park et al. [49] Sun et al. [50] Roselin et al. [51] CATComp (Proposed) between CoAP clients and servers via exchanging multicast messages. ...
... Hummen et al. [43] proposed Slimfit, a compression layer Mainly, the HIP header was the central focus. As we can see, this scheme does not consider payload for compression where our proposed module significantly address the compression of payload. ...
Full-text available
The Internet of Things (IoT) devices exchange certificates and authorization tokens over the IEEE 802.15.4 radio medium that supports a Maximum Transmission Unit (MTU) of 127 bytes. However, these credentials are significantly larger than the MTU and are therefore sent in a large number of fragments. As IoT devices are resource-constrained and battery-powered, there are considerable computations and communication overheads for fragment processing both on sender and receiver devices, which limit their ability to serve real-time requests. Moreover, the fragment processing operations increase energy consumption by CPUs and radio-transceivers, which results in shorter battery life. In this article, we propose CATComp -a compression-aware authorization protocol for Constrained Application Protocol (CoAP) and Datagram Transport Layer Security (DTLS) that enables IoT devices to exchange smallsized certificates and capability tokens over the IEEE 802.15.4 media. CATComp introduces additional messages in the CoAP and DTLS handshakes that allow communicating devices to negotiate a compression method, which devices use to reduce the credentials’ sizes before sending them over an IEEE 802.15.4 link. The decrease in the size of the security materials minimizes the total number of packet fragments, communication overheads for fragment delivery, fragment processing delays, and energy consumption. As such, devices can respond to requests faster and have longer battery life. We implement a prototype of CATComp on Contiki-enabled RE-Mote IoT devices and provide a performance analysis of CATComp. The experimental results show that communication latency and energy consumption are reduced when CATComp is integrated with CoAP and DTLS.
... Golomb Rulers provide the theoretical foundation to realize message dependencies such that the overlap between 0 5 10 15 20 25 (a) Golomb Ruler-based dependencies require the reception of more messages to achieve the full 128-bit security level than sliding window approaches, and thus increase the delay of integrity protection in a message stream, particularly for small tags (≤ 2 bytes). tags of neighboring messages is minimized. ...
... Not specifically focusing on message authentication, a final relevant branch of related work proposes to compress header fields of security protocols used in constrained environments such as DTLS [22,41,42] or HIP DEX [25]. Combining these approaches with our work allows further reduction to the communication overhead of security measures. ...
Message authentication guarantees the integrity of messages exchanged over untrusted channels. However, the required per-message authentication tags considerably expand packet sizes, which is especially problematic in constrained environments. To address this issue, progressive message authentication aggregates and distributes integrity protection over multiple messages, promising to reduce overheads while upholding strong security of traditional integrity protection. However, as we show in this paper, existing progressive message authentication schemes are susceptible to packet drops: By inferring with just two selected packets, an attacker can remove integrity protection from a complete sequence of messages. Revisiting the security of progressive message authentication, we consider it imperative to thwart such attacks by rethinking how authentication tags depend on the successful reception of packets. We propose R2-D2, which relies on (i) optimal message dependencies, (ii) parametrizable security guarantees, (iii) randomized bit dependencies, and (iv) optional immediate protection bits to address this problem. To deploy our approach to resource-constrained devices, we introduce SP-MAC, which implements R2-D2 using efficient XOR operations. Our evaluation shows that SP-MAC protects against sophisticated network-layer attacks and even operates more resource-conscious and faster than existing progressive message authentication schemes.
... If possible at all, modern security mechanisms have to be implemented in software only and run on the available hardware or be able to interface directly with the specialized devices used by grid operation companies, without affecting the availability of electricity supply. However, many devices in use may not have the computational power to support additional security functionality [50][51][52][53][54]. Even if certain devices are eventually exchanged or upgraded with new software, they are often required to support legacy protocols to be able to communicate with older devices still relying on these protocols. ...
Full-text available
Increasing volatilities within power transmission and distribution force power grid operators to amplify their use of communication infrastructure to monitor and control their grid. The resulting increase in communication creates a larger attack surface for malicious actors. Indeed, cyber attacks on power grids have already succeeded in causing temporary, large-scale blackouts in the recent past. In this paper, we analyze the communication infrastructure of power grids to derive resulting fundamental challenges of power grids with respect to cybersecurity. Based on these challenges, we identify a broad set of resulting attack vectors and attack scenarios that threaten the security of power grids. To address these challenges, we propose to rely on a defense-in-depth strategy, which encompasses measures for (i) device and application security, (ii) network security, and (iii) physical security, as well as (iv) policies, procedures, and awareness. For each of these categories, we distill and discuss a comprehensive set of state-of-the art approaches, as well as identify further opportunities to strengthen cybersecurity in interconnected power grids.
... If possible at all, modern security mechanisms have to be implemented in software only and run on the available hardware or be able to interface directly with the specialized devices used by grid operation companies, without affecting the availability of electricity supply. However, many devices in use may not have the computational power to support additional security functionality [41][42][43][44][45]. Even if certain devices are eventually exchanged or upgraded with new software, they are often required to support legacy protocols to be able to communicate with older devices still relying on these protocols. ...
Full-text available
Increasing volatilities within power transmission and distribution force power grid operators to amplify their use of communication infrastructure to monitor and control their grid. The resulting increase in communication creates a larger attack surface for malicious actors. Indeed, cyber attacks on power grids have already succeeded in causing temporary, large-scale blackouts in the recent past. In this paper, we analyze the communication infrastructure of power grids to derive resulting fundamental challenges of power grids with respect to cybersecurity. Based on these challenges, we identify a broad set of resulting attack vectors and attack scenarios that threaten the security of power grids. To address these challenges, we propose to rely on a defense-in-depth strategy, which encompasses measures for (i) device and application security, (ii) network security, (iii) physical security, as well as (iv) policies, procedures, and awareness. For each of these categories, we distill and discuss a comprehensive set of state-of-the art approaches, and identify further opportunities to strengthen cybersecurity in interconnected power grids.
... An interesting analysis of the use of Elliptic-curve cryptography in HIP is provided in Reference [18]. The Slimfit [19] protocol further optimized HIP-DEX by compression of the HIP header, which effectively reduced the fragmentation rate. The compression approach was combined with the distribution approach in the CD-HIP (Compressed and Distributed HIP) protocol [20], which targeted 6LoWPAN (IPv6 over Low-Power Wireless Personal Area Networks) communication. ...
Full-text available
The rapidly growing segment of the Internet of Things (IoT) makes the security threats more prominent than ever. The research around communication security and cybersecurity in such networks is still a challenge, mainly due to the typically limited energy and computation resources of IoT devices. The strong security mechanisms require significant power and thus the energy wastage must be minimized. Optimized application-specific security protocols are commonly used to make the data transfer more efficient, while still offering a high level of security. The supported security features, such as confidentiality, integrity or authenticity, should not be affected by the optimization. Our work is focused on optimizing one of the existing security protocols for the use in the IoT area, namely the Host Identity Protocol (HIP). Based on the analysis of related works, we have identified multiple possibilities for optimization and combined some of them into the proposed E-HIP optimized protocol. For verification purpose, it has been implemented as a modification of the open-source OpenHIP library and applied on a communication between real hardware devices. The secured communication worked correctly. The resulting effect of the proposed optimization has been evaluated experimentally and it represents an increase in energy efficiency by about 20%. Compared to other HIP optimizations, the achieved results are similar; however, the proposed optimizations are unique and can be further combined with some of the existing ones to achieve even higher efficiency.
Conference Paper
A huge amount of cybersecurity attacks discovered in recent years, moreover in the last one associated with the pandemic situation, where intruders are trying to exploit the hospitals and research centers. It can be dangerous as nowadays a lot of IoT devices are connected in that area of life. It is one of the several points of our motivation to secure IoT devices or data in Healthcare. In our article, we discussed several approaches dealing with the classification of security attacks. In the next part, we analyzed and compared HIP-based security protocols also used in IoT networks. We also ran experiments to show that it is crucial to bring energy-efficient lightweight, optimized protocol for using in IoT networks within sufficient security strength.
Full-text available
The goal of this work is to develop a key exchange solution for IPsec protocol, adapted to the restricted nature of the Internet of Things (IoT) components. With the emergence of IP-enabled wireless sensor networks (WSNs), the landscape of IoT is rapidly changing. Nevertheless, this technology has exacerbated the conventional security issues in WSNs, such as the key exchange problem. Therefore, Tiny Authenticated Key Exchange Protocol for IoT (TAKE-IoT) is proposed to solve this problem. The proposed TAKE-IoT is a secure, yet efficient, protocol that responds to several security requirements and withstands various types of known attacks. Moreover, TAKE-IoT aims to reduce computation costs using lightweight operations for the key generation. The proposed protocol is validated using the automated validation of internet security protocols and applications (AVISPA) tool. Hence, results show that TAKE-IoT can reach a proper level of security without sacrificing its efficiency in the context of IoT.
The maturity of the IoT depends on the security of communications and the protection of end-user's privacy. However, technological and material heterogeneities, and the asymmetric nature of communications between sensor nodes and ordinary Internet hosts, make the security in this case more problematic. Major problem facing the large deployment of IoT is the absence of a unified architecture and a lack of common agreement in defining protocols and standards for IoT parts. Many solutions have been proposed for the standardization of security concepts and protocols in IoT at different layers. Even though many advances and proposals were made for IoT adaptation as IPv6 for Low Power Wireless Personal Area Network (6LoWPAN), and at application layer with protocols such as XMPP, MQTT, CoAP, etc., security of the IoT remains a very challenging task and an open research topic. This chapter focuses on existing protocols and different proposed mechanisms in literature to secure communications in the IoT.
Conference Paper
Full-text available
Recent standardization efforts focus on a number of lightweight IP security protocol variants for end-to-end security in the Internet of Things (IoT), most notably DTLS, HIP DEX, and minimal IKEv2. These protocol variants commonly consider public-key-based cryptographic primitives in their protocol design for peer authentication and key agreement. In this paper, we identify several performance and security issues that originate from these public-key-based operations on resource-constrained IoT devices. To illustrate their impact, we additionally quantify these protocol limitations for HIP DEX. Most importantly, we find that public-key-based operations significantly hamper a peer's availability and response time during the protocol handshake. Hence, IP security protocols in the IoT must be tailored to reduce the need for expensive cryptographic operations, to protect resource-constrained peers against DoS attacks targeting these cryptographic operations, and to account for high message processing times. To this end, we present three complementary, lightweight protocol extensions for HIP DEX: i) a comprehensive session resumption mechanism, ii) a collaborative puzzle-based DoS protection mechanism, and iii) a refined retransmission mechanism. Our focus on common protocol functionality allows to generalize our proposed extensions to the wider scope of DTLS and IKE. Finally, our evaluation confirms the considerable achieved improvements at modest trade-offs.
Conference Paper
Full-text available
6LoWPAN is an IPv6 adaptation layer that defines mechanisms to make IP connectivity viable for tightly resource-constrained devices that communicate over low power, lossy links such as IEEE 802.15.4. It is expected to be used in a variety of scenarios ranging from home automation to industrial control systems. To support the transmission of IPv6 packets exceeding the maximum frame size of the link layer, 6LoWPAN defines a packet fragmentation mechanism. However, the best effort semantics for fragment transmissions, the lack of authentication at the 6LoWPAN layer, and the scarce memory resources of the networked devices render the design of the fragmentation mechanism vulnerable. In this paper, we provide a detailed security analysis of the 6LoWPAN fragmentation mechanism. We identify two attacks at the 6LoWPAN design-level that enable an attacker to (selectively) prevent correct packet reassembly on a target node at considerably low cost. Specifically, an attacker can mount our identified attacks by only sending a single protocol-compliant 6LoWPAN fragment. To counter these attacks, we propose two complementary, lightweight defense mechanisms, the content chaining scheme and the split buffer approach. Our evaluation shows the practicality of the identified attacks as well as the effectiveness of our proposed defense mechanisms at modest trade-offs.
Full-text available
Real deployments of the IoT require security. CoAP is being standardized as an application layer protocol for the Internet of Things (IoT). CoAP proposes to use DTLS to provide end-to-end security to protect the IoT. DTLS is a heavyweight protocol and its headers are too long to fit in a single IEEE802.15.4 MTU. 6LoWPAN provides header compression mechanisms to reduce the size of upper layer headers. 6LoWPAN header compression mechanisms can be used to compress the security headers as well. In this paper we propose 6LoWPAN header compression for DTLS. We link our compressed DTLS with the 6LoWPAN standard using standardized mechanisms. We show that our proposed DTLS compression significantly reduces the number of additional security bits. For example, only for the DTLS Record header that is added in every DTLS packet, the number of additional security bits can be reduced by 62%. Our compressed-DTLS is the first lightweight 6LoWPAN extension for DTLS.
This Recommendation provides cryptographic key management guidance. It consists of three parts. Part 1 provides general guidance and best practices for the management of cryptographic keying material. Part 2 provides guidance on policy and security planning requirements for U.S. government agencies. Finally, Part 3 provides guidance when using the cryptographic features of current systems.
Conference Paper
The vision of the Internet of Things considers smart objects in the physical world as first-class citizens of the digital world. Especially IP technology and RESTful web services on smart objects promise simple interactions with Internet services in the Web of Things, e.g., for building automation or in e-health scenarios. Peer authentication and secure data transmission are vital aspects in many of these scenarios to prevent leakage of personal information and harmful actuating tasks. While standard security solutions exist for traditional IP networks, the constraints of smart objects demand for more lightweight security mechanisms. Thus, the use of certificates for peer authentication is predominantly considered impracticable. In this paper, we investigate if this assumption is valid. To this end, we present preliminary overhead estimates for the certificate-based DTLS handshake and argue that certificates - with improvements to the handshake - are a viable method of authentication in many network scenarios. We propose three design ideas to reduce the overheads of the DTLS handshake. These ideas are based on (i) pre-validation, (ii) session resumption, and (iii) handshake delegation. We qualitatively analyze the expected overhead reductions and discuss their applicability.
This document is an informative overview of how legacy applications can be made to work with the Host Identity Protocol (HIP). HIP proposes to add a cryptographic name space for network stack names. From an application viewpoint, HIP-enabled systems support a new address family of host identifiers, but it may be a long time until such HIP-aware applications are widely deployed even if host systems are upgraded. This informational document discusses implementation and Application Programming Interface (API) issues relating to using HIP in situations in which the system is HIP-aware but the applications are not, and is intended to aid implementors and early adopters in thinking about and locally solving systems issues regarding the incremental deployment of HIP.
This document defines mobility and multihoming extensions to the Host Identity Protocol (HIP). Specifically, this document defines a general "LOCATOR" parameter for HIP messages that allows for a HIP host to notify peers about alternate addresses at which it may be reached. This document also defines elements of procedure for mobility of a HIP host -- the process by which a host dynamically changes the primary locator that it uses to receive packets. While the same LOCATOR parameter can also be used to support end-host multihoming, detailed procedures are left for further study.