tion 2 discusses about related work, whilst Section 3
describes the A-code primitive. In Section 4 we show
how to secure AODV thanks to A-coding. Section 5
sketch a comparison with the 802.15.4, and discusses
general A-code key management issues for AODV.
Finally, in Section 6 we draw out our conclusions, and
illustrate our future work.
2 RELATED WORK
Attacks against routing can be realized through a
suitable combination of offensive techniques such
as eavesdropping, identity spoofing, and the replay,
modification, forgery or deletion of routing control
packets. The adaptive nature of the communica-
tion mesh in ad-hoc networks, along with node’s
constrained resources, compel the adoption of on-
demand routing, which in case of wireless communi-
cations is particularly exposed to such offensive tech-
niques.
Many secure routing protocols for wireless ad-
hoc networks, based on both private-key and public-
key cryptography, have been proposed by the research
community over the last decade, with alternate re-
sults as in (Sanzgiri et al., 2002; Zapata and Asokan,
2002; Papadimitratos and Hass, 2003). Reporting on
such proposals is outside the scope of this work. A
comprehensive survey of secure on-demand routing
is given in (Hu and Perrig, 2004), whilst (Karlof and
Wagner, 2003) describes attacks on sensor network
routing protocols, and introduces some generic coun-
termeasures.
In virtually all cases, strategies have been adopted
to face against node’s constrained resources and
achieve acceptable overheads. Those approaches,
not surprisingly, have often resulted in some secu-
rity weakness or in assumptions that are difficult to
be satisfied in practice (Ramachandran and Yasinsac,
2001). For example, in (Sanzgiri et al., 2002; Hu
et al., 2003; Hu et al., 2005) authentication is real-
ized only for routing control packets, exposing data
packets to serious threats. Both SAODV (Zapata and
Asokan, 2002) and ARAN (Sanzgiri et al., 2002) pro-
vide message authenticity only when all intermediate
nodes are trustworthy, which is an overly restrictive
assumption. The TESLA authentication framework
(Perrig et al., 2000) - along with its variant µTESLA
(Perrig et al., 2002), specifically designed for wire-
less sensor networks, avoid hop-by-hop authentica-
tion by relying on loosely time synchronized network
nodes. However, secure time synchronization has
been demonstrated to be very difficult to achieve, also
on networks with a fixed infrastructure (Menezes, et
al.,1996). Consequently, all the routing protocols
based on such frameworks(e.g. the protocol in SPINS
(Perrig et al., 2002), SEAD (Hu et al., 2003), SEAR
(Zhao et al., 2008)) suffer the same drawback.
Considered together, the above works and expe-
riences seem to indicate that effective secure proto-
cols for ad-hoc routing can only be achieved if hop-
by-hop authentication is guaranteed for all the pack-
ets involved in the protocol. Both LHAP (Zhu et al.,
2003) and HEAP (Akbani et al., 2008) were designed
to offer hop-by-hop authentication for data packet as
well as control packets. LHAP realizes message au-
thentication through one-way hash key chains, and
that turns out in a low efficiency in terms of mem-
ory requirements, since long time communications re-
quires long chains. HEAP is a modified version of the
HMAC algorithm that uses two keys and seems very
efficient for multicast communications. HEAP was
designed to defend against attacks originating from
nodes that are not authenticated members of the net-
work (outsider attacks).
The A-coding approach offers hop-by-hop au-
thentication, too, and in the same security assump-
tions than HEAP. However, it has one main advan-
tage: since hop-by-hop authentication is realized at
the lowest layer of the protocol stack, any protocol
can get its own proper authentication service by just
being coupled with suitable key management, without
any modification to the protocol itself.
A-codes are based upon I-codes (Cagalj et al.,
2006) and, like these ones, were introduced to allow
for the establishment of authentic public keys over in-
secure radio channels (Schmid and Rossi, 2010). Dif-
ferently than I-codes, however, A-codes can provide
authentication without user intervention and in the ab-
sence of special, dedicated radio-frequency channels
(integrity channels).
3 THE A-CODE PRIMITIVE
The A-code primitive was introduced in (Schmid and
Rossi, 2010) to allow for the establishment of authen-
tic public keys in wireless sensor networks. It can
operate directly at the physical layer of the protocol
stack, on PPDUs
2
, offering a “physical coding” al-
ternative to traditional message authentication codes
(MACs).
In A-codes, message integrity is gained through
unidirectional message coding and on-off keying
communication with signal anti-blocking; these are
2
PPDU stands for PHY Protocol Data Unit, and repre-
sent the message structure managed by the IEEE Standard
802.15.4 (IEEE, 2006) physical (PHY) layer.
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
152