Caller Id Spoofing Detection

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

You Can Call But You Can’t Hide:

Detecting Caller ID Spoofing Attacks

Hossen Mustafa1 , Wenyuan Xu1 , Ahmad-Reza Sadeghi2 , and Steffen Schulz2,3


1
University of South Carolina
2
Technische Universität Darmstadt (CASED)
3
Ruhr-Universität Bochum, Germany & Macquarie University, Australia

Abstract. Caller ID (caller identification) is a service provided by tele-


phone carriers to transmit the phone number and/or the name of a caller
to a callee. Today, most people trust the caller ID information, and it
is increasingly used to authenticate customers (e.g., by banks or credit
card companies). However, with the proliferation of smartphones and
VoIP, it is easy to spoof caller ID by installing corresponding Apps on
smartphones or by using fake ID providers. As telephone networks are
fragmented between enterprises and countries, no mechanism is available
today to easily detect such spoofing attacks. This vulnerability has al-
ready been exploited with crucial consequences such as faking caller IDs
to emergency services (e.g., 9-1-1) or to commit fraud.
In this paper, we propose an end-to-end caller ID verification mecha-
nism CallerDec that works with existing combinations of landlines, cel-
lular and VoIP networks. CallerDec can be deployed at the liberty of
users, without any modification to the existing infrastructures. We im-
plemented our scheme as an App for Android-based phones and validated
the effectiveness of our solution in detecting spoofing attacks in various
scenarios.

1 Introduction
“What’s worse than a bad authentication system? A bad authentication
system that people have learned to trust” [1].

Caller ID services transmit the phone number and/or the name of a caller to
the recipient (callee) as caller ID intending to provide informed consent to the
callee before answering calls. However, Caller ID has been increasingly used to
authenticate the identities of callers, or to verify their physical locations in several
systems, ranging from 9-1-1 emergency services, automatic telephone banking
systems4 , credit card activation systems, to voicemail services. Unfortunately,
existing caller ID protocols do not provide real authentication and hence caller
IDs are vulnerable to spoofing attacks; i.e., an attacker can easily send a fake
caller ID to a callee. This vulnerability has already been exploited in variety of
4
For instance, Bank of America only requires a customer to enter a debit/credit card number to
access account information when the caller ID matches their records.
2

misuse and fraud incidents: In the US, thousands of people were victimized by
credit card fraud with the help of caller ID spoofing [2,3], causing a loss of more
than $15 million dollars annually; caller ID spoofing is also a common technique
used for swatting, which is an attempt to trick an emergency service with false
reporting of an incident — for instance, police officers were tied-up in responding
to a non-existent robbery reported by pranksters [4]; drugs were misused as a
result of spoofed pharmacists’ phone numbers [2]; other incidents include identity
theft, purchase scams [1], etc. Due to the proliferation of detrimental incidents
caused by caller ID spoofing, the US government passed the legislation Truth in
Caller ID Act of 2009 [5] making it illegal to transmit misleading or inaccurate
caller ID information with the intend to defraud.
However, today spoofing caller IDs has become much easier, because many
VoIP providers allow anyone to claim arbitrary caller IDs through VoIP client
software (e.g., x-lite [6]), and fake ID providers allow their customers to claim
any caller ID by simply dialing a special phone number or by utilizing readily
available Apps on smartphones (e.g., Caller ID Faker [7]). Thus, in this paper,
we focus on detecting caller ID spoofing attacks.
Caller ID spoofing is possible because caller IDs are transmitted in plaintext
with no authentication mechanisms in place. When a call is routed between
different carriers, the callee’s carrier will simply accept the caller ID claimed by
a caller’s carrier. Given the lack of authentication between carriers, caller IDs
could be trustworthy if (a) the telephone service providers do not manipulate
caller IDs, (b) the telephone infrastructure is tightly controlled, and no intruders
could tap into the infrastructure to create an arbitrary caller ID.
These conditions were true in the early days as the telephone network used
dedicated lines operated by a monopoly. Today, with current converging phone/data
networks and diversity of telephone service carriers, neither holds any more.
Moreover, telephone carriers may not be able to solve the problem even if they
are willing to redesign the protocols. This is because the entire telephone infras-
tructure comprises several telephone carriers with their own trust domains, and
a carrier can at most verify calls originated in its own network but not from
the other networks. To the best of our knowledge, no mechanism is currently
available to users for detecting caller ID spoofing without answering the call first
or without a special interface (and agreement) provided by the carrier.5
Challenges and contributions. We present an end-to-end detection mecha-
nism against caller ID spoofing attacks. Our approach utilizes a covert channel
between end users and does not require changing to the existing core telephone
networks. Such a detection mechanism is challenging to realize: First, only lim-
ited information and resources are available at end users. The route of call sig-
nalling is unknown. Second, compatibility to different protocols (GSM, VoIP,
PSTN) limits the design space. Third, any deviation from the regular calling
procedure is unlikely to be accepted by most people. Thus, naive solutions such
as rejecting an incoming call and then calling back, are not an option. The de-
5
A commercial proprietary service (TrustID) claims to detect caller ID spoofing [8] for business
customers. However, it is closed source and we were unable to obtain/analyze its technical solution.
3

tection mechanisms should be automated and require little user input. Fourth,
a few legitimate services provided by telephone companies allow the caller IDs
to be different from the calling numbers, making those caller IDs appear to be
spoofed. However, those scenarios should not be classified as caller ID spoofing
attacks. We address all these requirements and design an end-to-end caller ID
verification scheme which we call CallerDec. We summarize our contributions as
follows:
– We propose CallerDec, an end-to-end caller ID verification scheme that re-
quires no modification to the existing telephone infrastructure and is appli-
cable to calling parties using any telephone services. CallerDec can detect
spoofing even if a caller ID is not in the contact list or is unreachable.
– We present two use cases of CallerDec, one for an emergency call scenario
(e.g., 9-1-1 call) and the other for a regular call scenario. In both cases,
the end users, (e.g., a 9-1-1 service or an individual customer) can utilize
CallerDec to verify caller IDs.
– We implement CallerDec as an App for Android-based smartphones where
we tackle several technical challenges caused by the limited API support for
controlling calls. We examine the CallerDec performance in various scenarios,
and show that it can detect spoofed caller ID effectively and efficiently.

We stress that, while we implemented CallerDec on Android smartphones as


a case study, our solution can also be integrated in any other telephone devices.

2 Caller ID Spoofing Attacks


Creating a phone call typically involves two types of channels: a control channel
for signalling, and a voice channel for transmitting voice data. Caller IDs are
transmitted in control channels: When a caller dials a number, the carrier first
authenticates the caller, and then looks up the associated caller ID. Finally, the
caller ID is forwarded to the callee, possibly from one carrier to another.
It is difficult to spoof caller IDs by directly exploiting the Public Switched
Telephone Network (PSTN, landlines) or cellular network protocols because
caller IDs are automatically generated by landlines or cellular carriers, and con-
trol channels are not easily accessible to customers. However, it is easy to spoof
caller IDs in VoIP, since VoIP transmits both voice and control data in IP pack-
ets, and a caller can often set up any caller ID for an outgoing call. In addition,
the protocols that interconnect carriers, which include Signaling System No. 7
(SS7) [9] and VoIP, do not contain any caller ID verification mechanisms, and
a carrier will simply accept and forward the claimed caller IDs. Thus, spoofing
attacks require little effort in several ways. We discuss several attacks in the
following.
Spoofing via Fake ID Providers. Fake ID providers offer caller ID spoof-
ing services. They establish SS7/VoIP connections with various telephone carri-
ers (e.g., AVOICS [10]), and act as intermediary between attackers and victims
to relay caller IDs specified by its customers (attackers in this case). Fig. 1 illus-
trates an example, where an attacker (Eve) tries to call the victim (Bob) faking
4

Fig. 1: An illustration of how a fake ID provider spoofs a caller ID leveraging the


loophole in network interconnection protocols.

Alice’s caller ID. First, Eve calls the fake ID provider, and supplies Bob’s phone
number as the destination number and Alice’s phone number as the desired
spoofed caller ID. Then, the fake ID provider establishes a call to Bob with Al-
ice’s caller ID, and finally connects Eve with Bob once the call is answered. Eve
can subscribe to a fake ID provider and carry out spoofing attacks towards any
victim from any type of phone, provided that the fake ID provider is connected
to the victim’s network.
Spoofing via VoIP Services. Many VoIP carriers allow their customers to
specify their own caller ID, and will forward the caller ID to the callee’s carrier
without modifications. An adversary can subscribe to a VoIP carrier that allows
caller ID manipulation and can either use VoIP client software or a VoIP phone
to claim arbitrary caller IDs.
Spoofing via Automated Phone Systems. Automated phone systems
provide Interactive Voice Response (IVR) services for purposes of marketing,
survey collection, etc. Some service providers (e.g., Voxeo [11], Nuance Cafe [12])
allow their subscribers to select their own caller IDs and will deliver the selected
caller IDs for their subscribers regardless of their intention. Because these provi-
ders connect to major telephone carriers via SS7 or VoIP protocols [13], the
downstream telephone carriers will simply accept any caller IDs, including the
spoofed ones.
In this paper, we only evaluate our caller ID spoofing detection schemes
utilizing a fake ID provider. We believe that our proposed solution is capable
of detecting all aforementioned spoofing attacks since our detection scheme is
independent of how caller ID spoofing attacks are launched.

3 Assumptions and Requirements


3.1 System Model
For the rest of the paper, we will refer to Alice as the caller, Bob as the callee,
and Eve as the attacker who tries to spoof Alice’s caller ID while calling Bob. We
note that, Alice may not be in Bob’s contact list (unknown), and Alice’s number
could be invalid (unreachable). Since the verification operation is preformed
automatically, we expand our definition of the names and refer Alice, Bob, and
Eve to their devices as well. We envision that Alice, Bob, and Eve can be a
smartphone, a mobile phone, a PSTN phone, a VoIP phone, or an automated
5

(a) emergency call scenario (b) regular call scenario

Fig. 2: Two use cases of CallerDec.

system (e.g., bank), etc. Regardless of the type, we assume that Bob has a strong
incentive to verify the caller ID of a caller, e.g., he can be a bank that needs to
verify the caller ID of a customer. Thus, Bob integrates CallerDec in his device
(e.g., by installing an app in a smartphone, or by upgrading the firmware of a
PSTN phone, or by updating the software of a Private Branch Exchange (PBX)6 ,
etc). In comparison, Alice may or may not integrate CallerDec.
We assume that telephone carriers are trusted; they route outgoing calls
to dialed numbers and do not collude with Eve in any way. Thus, Eve cannot
capture or inject any type of packets into the telephone networks. Neither can
she answer or reject a call unless she is the callee. Additionally, we assume that
Alice does not collude with Eve and will not help Eve with caller ID validation.
Otherwise, we consider that Eve is authorized to use Alice’s caller ID.

3.2 Requirements
Security. The detection scheme should guarantee that an honest caller can prove
the validity of his/her caller ID, and an adversary cannot pretend to be calling
from an arbitrary number.
Compatibility. The detection solution should only change telephone terminals
but not the existing telephone infrastructure, and it should be compatible to
various telephone networks (e.g., GSM, VoIP, PSTN).
Usability. The detection strategies should be user-friendly, i.e., they should be
automated, require almost no effort from either a caller or a callee, and should
not change common procedures of phone calls. Otherwise, the callee could just
dial the displayed caller ID and verify verbally.
Efficiency. The detection scheme should have low computational overhead so
that it can be integrated into telephone terminals that have limited resources,
e.g., PSTN phones, mobile phones, etc.

3.3 Use Cases


CallerDec relies on a covert channel that is built on top of end-to-end communi-
cation services in telephone networks for verification. Overall, CallerDec works
as follows (Fig. 2): When Bob receives a phone call, CallerDec will automati-
cally initiate the caller ID verification by sending a challenge to Alice over the
6
Business organizations use PBX as phone exchanges which offer internal phones service, multiple
simultaneous calls with the same caller ID, etc.
6

end-to-end communication service. The challenge will be delivered to Alice if she


is reachable. Once the challenge reaches Alice, the CallerDec at Alice’s end will
respond to Bob with whether she has made the phone call. Collaboratively, the
CallerDec clients on both ends can automate caller ID verification. Even if Alice
is unreachable, Bob can still identify the situation and verify the caller ID.
We present two use cases of CallerDec: emergency call, where calls need to
be answered immediately, and regular call, where Caller IDs are expected to
be verified before calls are answered. In both cases, CallerDec follows the same
verification protocol, because verification relies on an end-to-end covert channel
that is independent to whether the original call is answered or not.
Emergency Calls. In emergency call cases, such as 9-1-1 services, caller
ID verification is performed in parallel to the voice call. As shown in Fig. 2(a),
the callee (e.g., an 9-1-1 service) answers the call from Alice upon ringing, and
CallerDec starts the verification process in the background. After the caller ID
is verified, CallerDec will notify the callee, and sensitive information may be
exchanged thereafter. Since the duration of 9-1-1 calls are reported to be between
1-2 minutes on average [14] [15], the verification results shall be returned before
the call is terminated.
Regular Calls. Eve may spoof Alice’s caller ID with the goal of winning a
chance talking to Bob, who would refuse otherwise (e.g., Eve may be an unknown
number or is on Bob’s block list). Thus, in regular call scenarios, CallerDec
performs the verification before a call is answered. As shown in Fig. 2(b), once
Bob receives an incoming call, CallerDec starts to verify the caller ID and notifies
Bob after the verification completes. While CallerDec may introduce delay in
answering phone calls, it allows users to answer or reject spoofed calls. We note
that a user may use CallerDec to verify only pre-selected numbers and balance
the trade off between delays and trust.

4 CallerDec: Verifying Caller IDs


4.1 Overview
The basic idea of CallerDec is to create a trusted covert channel between Alice
and Bob, i.e., the channel allows Alice and Bob to perform a challenge-response
but it is inaccessible to Eve. Forming such an end-to-end covert channel is diffi-
cult as CallerDec considers a telephone network as a black box and hence only
the services that are available to end systems can be used, e.g., Short Message
Service (SMS), traditional phone call, etc. Additionally, there are incompati-
bility and legacy issues in existing protocols. For instance, SMS is often not
available in PSTN carriers, which limits its use. To make CallerDec independent
of telephone networks, we utilize the traditional phone call service to form an
end-to-end covert channel between Alice and Bob.
Essentially, the covert channel is built on top of the control channel that is
used for call signalling in a traditional telephone network. Even though Alice and
Bob cannot manipulate control channels directly, they can acquire the status of
the phone call (e.g., answered/rejected). Since Eve cannot control or access the
7

Fig. 3: Call establishment and verification process: Alice is calling Bob who starts
v erification call after τsv interval, and Alice rejects the call after τv interval to prove
her caller ID.

calls between Alice and Bob, they form a trusted covert channel by initializing,
answering, or rejecting phone calls between them.
When Bob receives a call from Alice, he will initiate a new call to Alice
after a “starting verification” interval τsv and Alice will respond to the new call
according to whether she is indeed calling Bob. We refer to the first call from
o
Alice to Bob as the original call denoted by CA→B and the second call from Bob
v
to Alice as the verification call denoted by CB→A . Bob determines whether the
o
original call CA→B is indeed from Alice by examining the following information
that is sent over the control channel: (a) how Alice responds to the verification
v
call CB→A , and (b) how long Alice waits before responding. For instance, if
v
Alice is calling Bob, Bob will observe that she rejects the verification call CB→A
after a pre-defined interval τv . Because timing estimation of Alice’s waiting time
τv is performed at Bob’s side and its accuracy depends on the packet delivery
delays inside telephone networks, we use a probabilistic classifier to achieve high
estimation accuracy. As we discuss in Section 5, we use a Bayesian classifier [16]
that is suitable to resource constrained phone terminals.
We note that both τv and τsv are parameters used to differentiate whether
Alice supports CallerDec or not and should be the same in all CallerDec imple-
mentation. The values of τv and τsv should be chosen so that a user will unlikely
to respond to a call after a τv interval yet keeping the verification delay small.
The idea of forming a covert timing channel between Alice and Bob is sim-
ple. However, several open problems remain, such as: (a) Bob must be able to
estimate Alice’s waiting time τv at his end, (b) the protocol must handle all pos-
sible scenarios, i.e., caller ID is valid, caller ID is spoofed, Alice does not support
CallerDec, Bob’s verification call goes to Alice’s voicemail, etc. We address all
these issues in the design of CallerDec. In the following, we first discuss regular
call setup process in Section 4.2, then present CallerDec protocol in Section 4.3,
and perform security analysis in Section 4.4.
8

o
4.2 Regular Call Setup: CA→B
Without loss of generality, we consider the case that Alice and Bob belong to
carrier-A and carrier-B respectively, and the two carriers communicate through
SS7. We depict a regular call setup procedure in Fig. 3: when Alice dials Bob’s
number, a SETUP request is sent to carrier-A. Then, carrier-A sends carrier-B an
Initial Address Message(IAM), which is equivalent to SETUP. After carrier-B
sends a SETUP to Bob, he responds with an ALERTING message and starts the
ringing. The ALERTING message indicates that Bob is available and the ring-
ing has started. At this point, carrier-B sends carrier-A an Address Complete
Message(ACM). Subsequently, carrier-A sends Alice an ALERTING message, and
Alice starts to play the ringback tone.

4.3 CallerDec Verification Protocol


Continuing with the example shown in Fig. 3, we introduce CallerDec verification
protocol using the following scenarios: (a) normal scenario: Alice is indeed calling
Bob and both of them installed CallerDec, (b) attack scenario - spoof a reachable
user: Eve is spoofing Alice’s number, and Alice is reachable with CallerDec
installed, (c) attack scenario - spoof an unreachable user: Eve is spoofing Alice’s
number, and Alice is unreachable, and (d) not-supported scenario: Alice does not
install CallerDec. Depending on the type of use cases, Bob may require different
number of concurrent calls. For regular call cases, Bob only requires one call.
For emergency call cases, two concurrent calls are required since the verification
shall be performed in parallel to the original call.
o
Normal Scenario. As shown in Fig. 3, after receiving a phone call CA→B from
Alice, Bob will perform verification as follows.
v
1. After an interval T1 = τsv , Bob initiates a verification call CB→A to Alice
that triggers a sequence of six messages: SETUP, IAM, SETUP, ALERTING,
ACM, and ALERTING.
v
2. When Alice receives the verification call CB→A , she will reject it after an
interval T2 = τv . As a result, Bob will receive a REJECT message from carrier-
v
B indicating that CB→A has been rejected.
3. After receiving a REJECT message, Bob will measure the time difference T3
between the moment of sending the SETUP message and receiving the REJECT
message (Fig. 3). Examining T3 using the classifier, Bob will verify whether
Alice has waited for the expected time τv before rejecting verification call
v
CB→A . Once the waiting time is verified, Bob will further check the status
o o
of the original call CA→B . If CA→B is still active, Bob will conclude that
o
the caller ID is VALID. If CA→B is no longer active, then there is no need to
continue with the verification process.
As we discussed in Section 3.1, Eve cannot inject packets to the traditional
telephone networks, neither can she reject or answer the verification call directed
to Alice. Thus the verification process between Bob and Alice is protected, and
the response from Alice is trusted. We show a simplified version of this scenario
in Fig. 4(a).
9

(a) normal scenario

(b) spoof a reachable user


(c) spoof an unreachable user

Fig. 4: Simplified CallerDec protocol and outcomes in normal and attack scenarios.

Spoof a Reachable User. Eve is calling Bob and Alice is reachable, as shown
in Fig. 4(b). Similar to the normal scenario, Bob will first initiate a verification
v
call CB→A to Alice once he receives the call from Eve. Alice will treat Bob’s
v o
verification call CB→A as a regular call CB→A since she is not calling Bob. As
v
a result, instead of rejecting it, she will initiate a new verification call CA→B
after an interval τsv . When Bob identifies that Alice has initiated a verification
v
call CA→B , he concludes that Alice was not calling him. Instead, Alice is trying
v
to verify Bob’s verification call CB→A . After confirming that Alice’s verification
v
call CA→B was initiated after a duration of τsv , Bob will conclude that the caller
v o
ID is SPOOFED. He will terminate his verification call CB→A (CB→A for Alice)
v 7
and reject Alice’s verification call CA→B after an interval τv . Consequently,
v
Alice detects that her verification call CA→B has been rejected and the original
o
call CB→A she received is terminated. Then, she concludes that Bob may have
received a call that had spoofed her caller ID and terminates her own verification.

Spoof an Unreachable User. In this scenario, Eve is calling Bob, and Alice
is unreachable, e.g., her phone can be powered off, out of the coverage range, or
Alice is an invalid number. In such cases, as shown in Fig. 4(c), the verification
v
call from Bob CB→A will be directed immediately to either Alice’s voicemail or
v
carrier’s voicemail. When CB→A goes straight to voicemail, it contradicts to the
fact that “Alice” was calling Bob. Based on the timing estimation Bob can detect
that the verification call went straight to voicemail and will conclude that the
caller ID is SPOOFED.

Not-supported Scenario. Now, we discuss the case when Alice does not sup-
v
port CallerDec. In this case, the verification call CB→A will be considered as a
regular call. Since CallerDec is not installed, Alice (the person) may reject the
call after a random interval, answer the call, or even not respond to the call.
Regardless of the response, CallerDec can handle all cases:
(a) Normal Scenario. Without CallerDec, Alice may answer the verification call
from Bob. To leverage Alice’s knowledge, Bob’s CallerDec will play a pre-
7 v v
Bob rejects CA→B after τv to indicate that he did initiate the verification call CB→A .
10

initiate verification call

check call status


reject voicemail
new verification call answer

reject yes straight to


verification
after 2v SUHVVHG ³1´ voicemail
no call after 2sv
no yes
yes no no
yes

original no yes no
SUHVVHG ³2´
call active?

yes

VALID SPOOFED NOTSUPPORTED

end

Fig. 5: This flowchart shows how CallerDec handles Fig. 6: CallerDec on smart-
different cases to detect caller ID spoofing. A new phone for VALID and
incoming call initiates verification process. SPOOFED caller ID.

recorded voice instruction which asks Alice to press “1#” for confirming the
caller ID or to press “2#” to reject the verification. To proof her Caller ID,
Alice will press the proper keys, and Bob will conclude that the caller ID
is VALID. Alternatively, Alice may press a random key, and then Bob will
conclude that CallerDec is NOTSUPPORTED at Alice’s end. In addition, Alice
may ignore the call or may reject the call. For both responses, Bob will
conclude NOTSUPPORTED.
(b) Spoof a Reachable User. Similarly, Alice may answer the verification call.
After Alice enters the proper input (i.e., “2#”), Bob will conclude that the
caller ID is SPOOFED. For all other key-press (except “1#”), Bob will conclude
NOTSUPPORTED. In cases that Alice rejects the verification call, Bob will use
the classifier to verify whether Alice has waited for an interval τv before
rejecting the call. Since the value of τv is chosen beforehand to ensure that
it is unlikely for a human to reject calls after τv interval (e.g., τv = 0), Bob’s
CallerDec concludes NOTSUPPORTED. In cases that Alice ignores the call, Bob
will conservatively conclude that CallerDec is NOTSUPPORTED.
(c) Spoof an Unreachable User. The verification call will go to a voicemail, and
Bob can identify the situation utilizing the classifier and will conservatively
conclude that CallerDec is NOTSUPPORTED.

The overall decision process is illustrated in Fig. 5. We include NOTSUPPORTED


to address the cases when CallerDec is not installed. We envision that after
CallerDec is supported by all telephone devices, we can eliminate NOTSUPPORTED.

4.4 Discussion
Security Analysis. The security of this mechanism relies on the observation
that the verification call from Bob to Alice will be routed to Alice if she is
available and to a voicemail if she is unavailable, and Eve cannot manipulate
11

the verification call. Based on the choice of use cases, Bob can determine when
to answer a call, e.g., before the caller ID is verified or after. We stress that the
caller ID verification process is independent of when a call is answered. Hence,
regardless of the use cases (Section 3.3), Bob can utilize the same CallerDec to
verify caller ID.
In case of spoofing a reachable user (Section 4.3), equipped with CallerDec,
Alice will treat Bob’s verification call as a new call and will initiate a new veri-
fication call to Bob. Consequently, Bob concludes that the caller ID is SPOOFED
and Alice will conclude that Bob received a SPOOFED call. Without CallerDec,
when Alice receives the verification call from Bob, she may answer the call and
enter a proper input which leads Bob to conclude SPOOFED. If Alice rejects or
ignores the call, Bob will conservatively conclude NOTSUPPORTED, as discussed
in non-supported scenario(b) of Section 4.3. The bottom line is that Eve cannot
send any signal to convince Bob that Alice is calling.

Special Cases. (a) Blocked caller IDs. CallerDec depends on the caller ID
of an incoming call for verification and CallerDec cannot initiate verification
process if caller ID is BLOCKED or UNAVAILABLE. However, if Bob sup-
ports the mechanism to uncover caller ID of such calls, then CallerDec can be
integrated seamlessly. We note that 9-1-1 service has such capability, and if inte-
grated, CallerDec can perform caller ID verification effectively. (b) PBX systems.
CallerDec can be integrated easily in a PBX system of an organization, e.g., a
bank. Since such systems generally have resources for multiple concurrent calls,
they can adopt parallel verification, as discussed in use case 2 (Section 3.3).
Furthermore, if Alice is a PBX system and calls Bob, Bob can verify the caller
ID as usual. (c) Legitimate caller ID ‘spoofing’. It is possible that Alice inten-
tionally spoofs her own caller ID when calling Bob, e.g., Alice uses skype to call
Bob, while pretending to call from her cell phone. In this case, she can control
CallerDec on her cell phone, and thus can proof her identity. We consider this
scenario as a legitimate caller ID ‘spoofing,’ and CallerDec will conclude VALID.

Race Conditions. In a regular call scenario, both Alice and Bob may try to
call each other simultaneously. In such cases, both calls will go straight to the
voicemail. This is because most standards support call signalling for one active
call at a time, i.e., Alice or Bob cannot receive an incoming call while making an
outgoing one [17]. In this scenario, CallerDec is not initiated. In the case where
one of the calling party starts the call earlier than the other, one call will go
through at the best and CallerDec handles the case as usual.
When Eve tries to spoof both Alice and Bob simultaneously, both Alice and
Bob will initiate verification call to each other. But these calls would go straight
to the voicemail and CallerDec will correctly conclude SPOOFED.

5 Implementation and Validation


5.1 Implementation Challenges
When implementing CallerDec in Android, we encountered several challenges.
Particularly, CallerDec requires to automatically initiate a verification call, to
12

obtain the status of that call, and to estimate the ringing duration at the other
end. However, Android does not allow two concurrent phone calls and hides
the APIs for automating phone calls. Neither does Android contain APIs for
identifying the status of an outgoing phone call or estimating the ringing duration
at the other end. We discuss how we overcome these challenges in the following.

Initiate the Verification Call. Depending on the number of concurrent phone


calls, two categories of phone services exist: (a) primary rate interface (PRI) [18]
lines, and (b) regular lines (e.g., a mobile phone or a residential landline). PRI
supports multiple concurrent phone calls using the same caller ID. Thus, a sec-
ond line can be used to initiate the verification call while the first call may be
in progress. Regular end users can dial a secondary phone call but at most one
call can be active at a time. For instance, UMTS requires to put an incoming
call on hold before initiating a new call [17] and Android enforces this require-
ment. As a result, when implementing CallderDec in Android, we have to put the
incoming call from Alice on hold before initiating the verification call. Unfortu-
nately, Android provides no official APIs for putting a call on hold. To overcome
the problem, we leverage Android hidden APIs of ITelephony interface using
java reflection. We created an interface ITelephony in CallerDec App with the
package name set as com.android.internal.telephony and added the func-
tion definition from the original ITelephony interface with an empty body. As a
result, CallerDec is able to call the hidden functions from ITelephony at runtime
and can perform call control operations (e.g., initiate a new call).

Identify the Status of the Verification Call. CallerDec scheme requires


Bob to identify the status of the verification call, i.e., whether the call has been
answered, rejected, or directed to the voicemail. This task poses several chal-
lenges. Android does not allow users to access call signalling messages during
call setup. As a result, we cannot identify call status directly from call setup
messages, e.g., REJECT message. Neither does Android provide any API that re-
turns whether the callee’s phone is ringing or the call is answered. The status
of an outgoing call is always OFFHOOK8 . So, we seek alternatives to identify the
status of an outgoing call.
To identify the status of the verification call, we utilized system logs. Logs
of each Android app are printed in the system shell and CallerDec continuously
monitors real-time logs using Runtime APIs. In particular, CallerDec monitors
logs of three built-in system apps: CallNotifier, AudioService, and Ringer.
Once a DISCONNECT log is printed by CallNotifier, CallerDec concludes that
the verification call is rejected. To identify the answer or voicemail status, Caller-
Dec searches for an audioOn log entry from AudioService and a stopRing()
log entry from Ringer. To differentiate between answered and voicemail, Caller-
Dec can record voice data using the microphone and identify the patterns of
voicemail greeting using available tools [19]. If the pattern matches, CallerDec
has reached the voicemail, otherwise the verification call is answered.
8
OFFHOOK traditionally indicates that the handset of a PSTN phone is off the base and the user
could be dialing a number or on an active call. It is used in the same context in Android.
13

12 12 12
DCSD DCSD DCSD
10 DCFD 10 DCFD 10 DCFD
Seconds

Seconds

Seconds
8 SCSD 8 SCSD 8 SCSD
6 SCFD 6 SCFD 6 SCFD
4 4 4
2 2 2
0 0 0
08 11 14 17 20 23 08 11 14 17 20 23 08 11 14 17 20 23
Time of day Time of day Time of day
(a) VALID (b) SPOOFED (reachable) (c) SPOOFED (unreachable)

Fig. 7: End-to-end verification delay in (a) the normal scenario when caller ID is
VALID, and in the attack scenarios when caller ID is SPOOFED with Alice is (b)
reachable and (c) unreachable.

Verify Caller ID Using Timing Estimation. Another key issue to verify


Alice’s caller ID is to estimate her ringing duration (denoted by T2 ). As shown
in Fig. 3, the ringing duration (T2 ) is the time difference between the moments
when Alice sends an ALERTING message, and a REJECT message.
We found that T4 (Fig. 3), which is the time difference between the moment
when Bob receives an ALERTING and the one when he receives an ANSWER or
REJECT message, is unable to estimate T2 because some carriers start playing
the ringback tone before receiving an ALERTING message. For instance, AT&T
starts the ringback tone even when the callee is unavailable.
We found that T3 (as shown in Fig. 3), which is roughly the sum of T2 and
the round trip time from Bob to Alice, is independent of the types of carriers,
since it does not depend on when the ringback tone starts. Hence we chose T3 for
estimating T2 . To make CallerDec compatible to devices with low computational
power, e.g., mobile phones, we choose Bayesian Classifier [16]. Bayesian clas-
sifier is an efficient method for calculating posterior probability based on prior
probability and likelihood in the training data. Although the classifier needs
prior training, our experimental results suggest that the same trained model can
be used on different phones for effective classification.
For the training dataset, we recorded the values of T2 , T3 , time of day (Tday ),
and status of the verification call (Scall ), i.e., rejected, answered or voicemail.
We label each dataset with appropriate class: VALID, SPOOFED, or NOTSUPPORTED.
For each test sample, we employ the following Bayes equation [16] to calculate
the probability of each class, Ci .

p(T3 , Tday , Scall Ci ) p(Ci )
p(Ci T3 , Tday , Scall ) = (1)
p(T3 , Tday , Scall )

Here, p(Ci ) is the probability of Ci in the training dataset. CallerDec classifies


the test sample as the class with the highest probability. Thus, based on the
estimated duration of T2 and Alice’s action, CallerDec detects caller ID spoofing
attacks.
Fig. 6 shows two screenshots of CallerDec when the caller ID is VALID and
SPOOFED. To save space, we omit the screenshot of NOTSUPPORTED.
14
15 VALID
SPOOFED (reachable)
SPOOFED (unreachable)

Seconds
Device Name Processor RAM Class 10
Google Nexus One 1 GHz 512 MB Fast
5
HTC Sense 1 GHz 576 MB Fast
MyTouch 528MHz 192 MB Slow
0
S. Carolina California Michigan Washington
State of the callee
Table 1: Configurations of Android Fig. 8: End-to-end verification delay of
devices used in performance analysis CallerDec based on geographic locations.

5.2 Performance

To evaluate the performance of CallerDec, we measured time of day and end-to-


end delay of completing caller ID verification for the following scenarios which
we discussed in Section 4.3: (a) normal, (b) spoof a reachable user, (c) spoof
an unreachable user, and (d) not-supported scenarios. Additionally, we studied
the impact of the type of phones, the carriers, and the time of the day in the
verification delay. We selected three Android devices and classified them as fast
devices or slow devices based on the configurations, and the device specifications
are summarized in Table 1. We chose some common telephone carriers in the
USA, which are AT&T, T-Mobile, and SimpleMobile. We used two cases in the
experimental setup: (a) Alice and Bob belong to the same carrier, e.g., T-Mobile,
and (b) Alice and Bob belong to different carriers, e.g., a T-Mobile user calls an
AT&T user. In total, we measured data at six different times of the day in four
experimental setup: (a) DCFD: Different Carriers and using two Fast Devices,
(b) DCSD: Different Carriers and using one fast and one Slow Device, (c) SCFD:
The Same Carrier and using two Fast Devices, and (d) SCSD: The Same Carrier
and using one fast and one Slow Device. We set τv = τsv = 0 seconds in our
implementation to minimize verification delay. Note that other threshold values
can be used depending on the network parameters.
End-to-end Verification Delay. We measure end-to-end verification delay as
the time difference between the moment when Bob receives an incoming call and
the one when he identifies Alice’s action. In the normal scenario, when caller ID
was valid (Fig. 7(a)), the verification was done in 8.40 seconds on average. In
the worst case, when the caller and the callee were under different carriers, and
one of them was using a slow device, the delay was 8.61 seconds. The call setup
delay dominates the delays. For instance, a recent study reported that call setup
in 3G networks is between 4-7 seconds on average for various scenarios [20].
In the spoofing a reachable user scenario, Alice initiated a verification call
after τsv seconds in response to Bob’s verification call. As shown in Fig. 7(b), the
verification was done in 8.35 seconds on average. Similar to regular scenarios,
the call setup delay dominates the end-to-end delay.
In the spoofing an unreachable user scenario, Alice’s phone was turned-off
and the verification call went straight to the voicemail. As shown in Fig. 7(c),
the verification delay was less than 2 seconds on average and 2.13 seconds in the
worst case. Note that verification delay is low in this scenario because the call is
not routed to its destination since the caller is unreachable.
15

1 1 1
0.9 0.9 0.9
Accuracy

Precision

Recall
0.8 0.8 0.8

0.7 0.7 0.7


VALID VALID
0.6 0.6 SPOOFED 0.6 SPOOFED
NOTSUPPORTED NOTSUPPORTED
0.5 0.5 0.5
0 50 100 0 50 100 0 50 100
Training data (%) Training data (%) Training data (%)

(a) accuracy (b) precision (c) recall


Fig. 9: Performance of our Bayesian spoof detection classifier where (a) shows the
accuracy, (b) shows the precision and (c) shows the recall of the 1classifier.
We also analyzed the latency of CallerDec based on the geographic locations
of the caller and the callee (Fig. 8). In our experiments, the caller was always in
South Carolina and the callee was in one of the four states: California, Michigan,
South Carolina, and Washington. The result indicates that geographic locations
of the caller and the callee have minor effects on the delay.
Although CallerDec takes a few seconds for end-to-end verification, our anal-
ysis shows that such delay is mainly caused by telephone networks, and end
devices or network loads have minor effects. However, the verification delay can
be hidden in case of emergency calls (Fig. 2(a)) because the verification is done
in parallel to the phone call. Although CallerDec adds delay overhead before a
user may answer calls in case of regular calls (Fig. 2(b)), the actually experienced
overhead should be lower since it generally takes a few seconds to answer a call.
Timing Estimation. To verify caller ID, Bob estimates Alice’s waiting time
τv using a Bayesian classifier to decide whether a call is VALID, SPOOFED or
NOTSUPPORTED. To analyze the performance of our classifier, we collected more
than 2800 instances of calls labelled with appropriate class, e.g., approximately
1100 VALID, 1100 SPOOFED, and 600 NOTSUPPORTED instances. For NOTSUPPORTED
class, Alice rejected or answered the verification call at random time. We divided
the dataset into training and test sets at various proportions p, where p = [0.1
- 0.9]. For instance, with p = 0.1, 10% (approximately 280 instances) of the
dataset was used for training and the rest 90% (approximately 2600 instances)
was used for testing. We use the following metrics for evaluating CallerDec clas-
sifier where 100% is the desired outcomes for each metric: (a) an accuracy which
is the percentage of correct outcomes of CallerDec, (b) a precision which is the
percentage of correct outcome for a class out of all CallerDec outcomes for that
class (e.g., correct VALID outcomes out of all VALID outcomes), and (c) a recall
which is the percentage of correct outcome out of all correct outcomes for a class.
As depicted in Fig. 9(a), the accuracy of the classifier is more than 99% even
when the percentage of the training dataset is only 10%, and 99.26% on average.
Furthermore, the precision and recall are fairly constant: a 99.98% precision
and a 98.91% recall when caller ID is VALID, a 100% precision and recall when
caller ID is SPOOFED, and 95.62% precision and 99.93% recall when CallerDec is
NOTSUPPORTED. The results also suggest that a small number of training data is
sufficient for efficient classification.
In summery, CallerDec can be used effectively to detect caller ID spoofing.
It provides high accuracy in caller ID spoofing detection.
16

6 Related Work
While the problem of caller ID spoofing is generally known, previous solutions
typically require the cooperation and modification of phone provider networks.
For instance, Cai [21] proposes several ways to validate caller ID information
based on the available meta-data of a call. However, the scheme does not cope
with fake ID providers, which can fake most of this meta-data. Additionally,
customers must rely on their respective phone providers to verify the claimed
caller ID. In the RealName Registry [22], phone providers establish authenti-
cated name registries within their respective jurisdictions. Customers are issued
cryptographic certificates by their providers and can verify each others’ caller
IDs. Unfortunately, the cost of globally upgrading equipment for cryptographic
authentication and PKI is prohibitive, and providers that sell fake caller IDs
may still provide spoofed certificates.
PinDrop [23] evaluates audio artifacts introduced by digital encoding and
analog interference. As the call is routed through the network, the different
deployed types of network technology succinctly manipulate the audio signal,
creating a characteristic watermark that can be used to reconstruct and recognize
the path taken by a call. Similar to our approach, PinDrop does not require
cooperation of the network providers and can be realized on an on-demand basis,
by unilaterally modifying the callee’s device software. However, PinDrop focuses
on detecting whether a known caller ID originates from an unusual network
location. Instead, we focus on detecting any caller IDs, including previously
unseen callers. We believe CallerDec and PinDrop can complement each other.
Piotrowski et al. [24] consider voice spoofing as an extension of caller ID
spoofing and propose a watermarking mechanism to mitigate the threat. How-
ever, their approach requires modification of the caller and callee’s devices. In
this case, when arbitrary modifications to the caller and callee can be assumed,
well-known cryptographic approaches can be employed (e.g., CryptoPhone [25]).

7 Conclusion
In this paper, we investigated caller ID spoofing attacks and designed an end-to-
end solution, which we call CallerDec, to detect a spoofed caller ID. CallerDec
verifies the caller ID using a covert channel, which is built on top of the ver-
ification call from the callee to the claimed caller, and CallerDec uses timing
estimation together with the call status for verification. We implemented Caller-
Dec in Android-based phones and validated that CallerDec can effectively verify
caller ID. Although the end-to-end delay for completing a verification takes a few
seconds, such delay can be hidden when the verification is performed in parallel
to the voice call.
We studied CallerDec on Android-based phones as a case study, but Caller-
Dec can be integrated to other types of phone terminals to protect end users
from caller ID spoofing attacks. In addition, the current CallerDec will conclude
NOTSUPPORTED when CallerDec is not implemented by a phone terminal. We en-
vision that NOTSUPPORTED can be eliminated once the CallerDec is supported on
all telephone terminals.
17

References
1. Schneier, B. http://www.schneier.com/blog/archives/2006/03/caller_id_spoof.html
2. Rep. Engel Anti-Spoofing Bill Passes House. http://engel.house.gov/latest-
news1/rep-engel-anti-spoofing-bill-passes-house
3. ABCNews: Caller ID Scam Solicits Personal Info, Money.
abcnews.go.com/GMA/Consumer/story?id=3305916 (2007)
4. Cuellar, D.: Pranksters Terrorize Delco Family in “swatting” Call. WPVI-TV,
Philadelphia, PA (2010)
5. US Congress: Truth in Caller ID Act of 2009. www.gpo.gov
6. X-Lite. www.counterpath.com/x-lite.html
7. Caller ID Faker-Fake a Call! www.calleridfaker.com
8. TrustID. www.trustid.com
9. ITU-T: Q.700. www.itu.int/rec/T-REC-Q.700-199303-I/en
10. AT&T: Voice Networking Solutions. www.business.att.com
11. Voxeo: Prophecy IVR Platform Software. www.voxeo.com
12. Bevocal Cafe: Supercharge Your Portal. cafe.bevocal.com
13. Eisenzopf, J.: What You Need To Know About Voice ASPs. www.datamation.com
14. Lincoln Emergency Communications Center Annual Report. www.lincoln.ne.gov
15. Dauphin County Emergency Management Agency Year Yearly Statistics.
www.dauphincounty.org (2011)
16. Han, J., Kamber, M.: Data Mining: Concepts and Techniques. Elsevier Inc (2006)
17. IETF: ETSI TS 122 083. www.etsi.org
18. ITU: Q.431 Primary rate interface. www.itu.int/rec/T-REC-I.431-199303-I/en
19. Carnegie Mellon University: CMU Sphinx. cmusphinx.sourceforge.net
20. QUALCOMM: Circuit-switched fallback. the first phase of voice evolution for
mobile LTE devices. Technical report, www.qualcomm.com (2012)
21. Cai, Y.: Patent Application: Validating Caller ID Information to Protect Against
Caller ID Spoofing. (2008)
22. Chow, S.T., Gustave, C., Vinokurov, D.: Authenticating displayed names in tele-
phony. Bell Labs Journal (2009)
23. Balasubramaniyan, V.A., Poonawalla, A., Ahamad, M., Hunter, M.T., Traynor,
P.: Pindr0p: Using single-ended audio features to determine call provenance. In:
CCS. (2010)
24. Piotrowski, Z., Gajewski, P.: Voice spoofing as an impersonation attack and the
way of protection. Journal of Information Assurance and Security (2007)
25. GSMK Crypto Phone. www.cryptophone.de
26. ETSI: UMTS. www.etsi.org
27. ETSI: W-CDMA. www.etsi.org
28. 3GPP: TS 24.081. www.quintillion.co.jp/3GPP/Specs/
29. Niemi, V., Nyberg, K.: UMTS Security. John Wiley & Sons (2003)
30. Biryukov, A., Shamir, A., Wagner, D.: Real time cryptanalysis of A5/1 on a PC.
In: FSE. (2000)
31. Meyer, U., Wetzel, S.: A Man-in-the-Middle Attack on UMTS. In WiSe (2004)
32. Livengood, D., Lin, J., Vaishnav, C.: Public switched telephone networks: A net-
work analysis of emerging networks. Technical report, mit.edu (2006)
33. Bell Communication Research: Bellcore Technical Specification.
www.morehouse.org/hin/blckcrwl/telcom/callerid.txt
34. British Telecomm: SIN 227. www.btwebworld.com
35. Hersent, O., Gurle, D., Petit, J.P.: IP telephony: packet-based multimedia com-
munications systems. Addison-Wesley (2000)
18

Fig. 10: An example telephone network architecture, where different carriers are connected using
peering architecture.

36. IETF: SIP: Session Initiation Protocol. www.ietf.org/rfc/rfc3261.txt


37. ITU: H.323 : Packet-based multimedia communications systems.
www.itu.int/rec/T-REC-H.323/e
38. Ono, K., Tachimoto, S.: Sip signaling security for end-to-end communication. In:
APCC. (2003)
39. SipDroid. code.google.com/p/sipdroid/
40. Ellig, J.: Regulatory Status of VoIP in the Post-Brand X World. Bepress Legal
Series (2006)
41. North American Numbering Plan Administration. www.nanpa.com
42. Berg, R.V.D.: The Future of Interconnection. 9th Global Symposium for Regula-
tors (GSR). (2009)

A Background
Three categories of telephone carriers are in service: cellular networks, Public
Switched Telephone Network (PSTN), and Voice over Internet Protocol (VoIP)
providers. In the following, we give an overview of the popular caller ID standards
used within each type of carrier and between different carriers with the goal of
understanding the feasibility of injecting spoofed caller IDs.
A.1 Cellular Network
Architecture. Universal Mobile Telecommunications System (UMTS) [26] and
Wide-band Code-Division Multiple Access (W-CDMA) [27] are the two most
popular technologies for providing cellular telephone services. Despite which
technology is used, a cellular telephone network follows a hierarchical structure.
As illustrated in Fig. 10, the simplest cellular network consists of the follow-
ing entities for voice services (from the top to bottom levels): Mobile Switching
Centers (MSC), Base Station Controllers (BSC), and Base Transceiver Stations
(BTS). The HLR stores all necessary data for caller ID services, authentication
and billing purposes and interacts with MSC directly.
Each mobile station (MS) has a Subscriber Identity Module (SIM) and a
mobile carrier authenticates an MS based on the SIM information. When an MS
makes a phone call, the call setup process always goes through BTS, BSC, and
MSC. Then, the MSC obtains the caller ID associated with the MS from the
HLR and encodes it in a control packet for call setup.
19

Protocols. In UMTS and W-CDMA, the caller ID is encoded in call setup


packets (in the Calling Party BCD Number field) using the Binary Coded Dec-
imal (BCD) format and has a variable length of 3-14 bytes [28]. It is possible
that a call is set up without caller ID, and the Presentation Indicator field
is used to indicate whether caller ID is present in the packet.

Feasibility of Caller ID Spoofing. 3GPP specification has security mech-


anisms which include MS authentication, random session keys and SIM secu-
rity [29]. Although the communication between MS and BTS is encrypted with
a session key, cracking the session key [30] and man-in-the-middle attacks [31]
have been reported. An attacker can take advantage of such vulnerabilities to
spoof caller IDs. However, to the best of our knowledge, no such caller ID spoof-
ing attacks have been reported.

A.2 Public Switched Telephone Network


Architecture. The PSTN is a circuit-switched telephone network, known as
landline telephone. The PSTN generally has a hierarchical architecture [32] with
Central Exchanges (CEs) at the top level of the hierarchy (Fig. 10). Local Ex-
changes (LEs) consist of several PSTN switches and one switch port is assigned
to one customer configured with caller ID. When a customer dials a number, the
LE sends the pre-configured caller ID in the outgoing call.

Protocols. There are several caller ID standards in PSTN, e.g., Bellcore FSK,
SIN227, DTMF, V23, ETSI FSK, etc. Among these, Bellcore [33] and SIN227 [34]
are the most popular. Most of the protocols use Frequency Shift Keying (FSK)
for transmission and transmit caller ID in plain text.

Feasibility of Caller ID Spoofing. All other standards transmit the caller


ID information in plaintext. However, it is not easy to launch the attack in-
side PSTN because the caller ID signal is generated automatically by the LE,
based on the pre-configured information. Such information cannot be changed by
unauthorized entities, because the switches (LEs) are generally kept in secured
cabinets, inaccessible to the general public.

A.3 Voice over Internet Protocol


Architecture. VoIP technology takes advantage of IP where both voice and
control data are transmitted in IP packets. Unlike PSTN or cellular networks,
VoIP usually follows a flat/p2p architecture [35]. The control channel always
follows the client-server model but the voice channel between a caller and a
callee may use a direct connection. For the call setup, VoIP uses protocols such
as Session Initiation Protocol (SIP) [36] or H.323 [37].

Protocols. Both SIP and H.323 have built-in support for caller ID. SIP uses
the From field in the INVITE packet to send the caller ID, and the caller ID can
be any ASCII characters with an arbitrary length. For instance, the caller ID
in SIP has the form sip:callerID@ip_address and is typically encapsulated
20

in SIP packets in plaintext. Although secured SIP is available to encrypt caller


IDs, they are not authenticated [38]. Similar to SIP, caller ID is also transmitted
in plaintext in H.323.

Feasibility of Caller ID Spoofing. Unlike PSTN or cellular network sce-


narios, in VoIP the caller ID originates at the client end; i.e., the clients could
generate control packets with arbitrarily chosen caller IDs. Most VoIP software
provides an interface allowing a caller to specify his/her caller ID for each phone
call, making caller ID spoofing trivial; e.g. x-lite [6], sipdroid [39], etc. In fact,
many VoIP carriers manipulate caller ID to avoid long distance charge [40].

A.4 Network Interconnection Protocols:


Different telephone networks are interconnected using a peering architecture, as
shown in Fig. 10. In the following, we discuss how a call is routed and caller ID
is forwarded between carriers.

Call Routing. In telephone systems, phone numbers are assigned based on


geographic locations to discriminate between local and long-distance calls. In
the US, each carrier is assigned a unique prefix in each geographic location by
the North American Numbering Plan Administration (NANPA) [41] and the
call routing is done based on prefix-matching. For example, in Washington, the
360-269 prefix is assigned to AT&T and 360-270 to Sprint. When a customer
calls 360-269-XXXX, the originating carrier must forward the call to AT&T9 .

Caller ID Forwarding. Signaling System No.7 (SS7) [9] is the de facto stan-
dard for interconnecting carriers, even though many regulators have suggested
using VoIP [42]. When a caller and a callee have subscribed to different carriers,
the call has to go through either an SS7 or VoIP connection. The originating
carrier sends the caller ID as part of the control packets. In both cases, the re-
ceiving carrier passes the caller ID data to the callee without any modification
or validation. Hence, caller ID is not verified in either case.

Feasibility of Caller ID Spoofing. Since there is no verification mechanisms


between carriers, it is possible for an attacker to get connected with a carrier
using either SS7 or VoIP, and then exploit the lack of authentication between
carrier networks to spoof caller ID. Such an attack, while valid, is costly and
complex to carry out because the attacker has to establish an SS7/VoIP connec-
tion with a carrier, which requires the attacker to complete an interconnection
agreement with the carrier, to install necessary hardware and software, to pay a
premium, etc. Thus, it is not meant for casual adversaries with a limited budget.

Summary. It is difficult to launch caller ID spoofing attacks by exploiting


the caller ID protocol within PSTN or cellular networks, but it is possible by
exploiting VoIP. Additionally, an adversary can spoof caller ID by exploiting the
lack of caller ID authentication between carriers.
9
Unless the phone number is ported, i.e., a customer of carrier A switches service to carrier B but
keeps the same phone number. In this case, the call will be routed to the carrier B instead of A.

You might also like