divendres, 31 de juliol de 2009

Com piratejar el teléfons mòbils amb uns "simple"SMS?

Amb varis SMS és possible bloquejar o fer trucades.

Els investigadors en informàtica Charlie Miller i Collin Mulliner volen explicar a tohom al salò Black Hat de La Vegas com o han fet.
El sistema funciona en mòbils amb sistemes operatius Windows mòbil, Android i el iPhone´s.

Comenten que si no expliquen el sistema qualsevol o pot utilitzar "discrètament".

La tècnica utilitza el SMS d´una manera força interessant.
Cada SMS envia un CÒDIC "dolent". El primer, un missatje per "despistar" i això activa l´acciò pirata. Després sen envien uns altres amb uns 100 caràcters cadascun.

Segons la versió dels dos informàtics l´ùnica sol.luciò és parar el telèfon!.

Doc original :
Fuzzing the Phone in your Phone
(Black Hat USA 2009)
Collin Mulliner
TU-Berlin / T-Labs
Charlie Miller
Independent Security Evaluators
June 25, 2009
In this talk we show how to find vulnerabilities in smart phones.
Not in the browser or mail client or any software you could find on a
desktop, but rather in the phone specific software. We present techniques
which allow a researcher to inject SMS messages into iPhone,
Android, and Windows Mobile devices. This method does not use the
carrier and so is free (and invisible to the carrier). We show how to
use the Sulley fuzzing framework to generate fuzzed SMS messages
for the smart phones as well as ways to monitor the software under
stress. Finally, we present the results of this fuzzing and discuss their
impact on smart phones and cellular security.
1 Introduction
The Short Message Service (SMS) is the oldest service existing in addition
to the actual mobile telephony service. SMS is loved by the users because
it is simple and easy. You punch in a short text message such as what’s
up dude? and hit send. The text almost instantaneously appears on the
receivers phone, and thats it from the user’s perspective. It is simple and
easy. For the mobile phone operator SMS is a good additional business
since the messages are relatively expensive and SMS is counted as unreliable
therefore messages can be delayed or lost without any repercussions. Also in
practice SMS seems pretty reliable.
Under the hood the Short Message Service is much more complex then
the user expects. SMS is used for all kinds of other mobile phone related
services such as voice mail notification, the Wireless Application Protocol
(WAP), multimedia messaging (MMS - the Multimedia Messaging Service),
Over-the-Air (OTA) remote phone configuration, and for vendor specific features
like the iPhone’s visual voice mail. To incorporate all these features the
Short Message Service supports sending binary messages that contain additional
control information, therefore making the SMS system very complex.
Complexity often leads to implementation faults. Implementation faults often
lead to security issues.
In the past years we have seen many different SMS-related security problems
on mobile phones. The problems range from crashing and rebooting
devices to bugs that prevent further reception of SMS messages. From a
security stand point SMS is the worst possible case since it is an always on
technology. As long as a mobile phone is connected to the mobile phone network
SMS messages can be delivered to the device. Further it is uncommon
for mobile phone service operators to filter SMS messages on the network.
Also since SMS is an essential component of the mobile phone service it cannot
be deactivated on a mobile phone. Even if the possibility existed users
would most likely never deactivate it, therefore, making SMS the perfect attack
vector against mobile and smart phones.
Until now most of the SMS related security issues have been found by
accident. People trying to use a specific feature on a specific phone or trying
to implementing some SMS-based application and in the process discover a
bug or vulnerability. One reason why no systematic analysis of SMS implementations
was conducted until now is the fact that sending SMS messages
costs money. Because of the lack of access to source code of SMS implementations
all testing has to take the black-box approach therefore requiring a
huge amount of SMS messages to be sent.
In this work we present a novel way for conducting vulnerability analysis
of SMS implementations that does not require sending SMS messages using a
mobile operator network. For our approach we inject SMS messages locally
into our test phones. The injection is done in software only and requires
only application level access to the phone. We inject SMS messages below
the software telephony stack and therefore are able to analyze and test all
SMS-based services that are implemented on the mobile telephony stack on
the respective phones.
The vulnerability analysis itself was conducted using fuzzing. In this
paper we present the possibilities for fuzzing-based testing of SMS implementations.
We further present our testing tools and test methodology.
2 The Short Message Service
The Short Message Service is a store and forward system, messages sent to
and from a mobile phone are first sent to an intermediate component in the
mobile phone operators network. This component is called the Short Message
Service Center (SMSC). After receiving a message, the SMSC forwards the
message to another SMSC or if the receiving phone is handled by the same
SMSC, it delivers the message to the recipient without invoking another
The SMSC can receive messages from all kinds of sources besides a mobile
phone for example from the voice mail system that wishes to inform the
mobile phone about a waiting voice mail message.
2.1 The SMS Message Format
SMS messages exist in two formats: SMS SUBMIT and SMS DELIVER.
The SMS SUBMIT format is used for messages sent from a mobile phone
to the SMSC. The SMS DELIVER format is used for messages sent from
the SMSC to the mobile phone. Since our testing method is based on local
message injection that replicates an incoming message, we are only interested
in the SMS DELIVER format.
2.1.1 The SMS DELIVER Format
An SMS DELIVER message consists of the fields shown in Table 1. The
format is simplified since our main fuzzing targets are the Protocol ID, the
Data Coding Scheme, and the User Data fields. Other fields such as the User
Name Bytes Purpose
SMSC variable SMSC address
DELIVER 1 Message flags
Sender variable Sender address
PID 1 Protocol ID
DCS 1 Data Coding Scheme
SCTS 7 Time Stamp
UDL 1 User Data Length
UD variable User Data
Table 1: SMS DELIVER Message Format
Field Bytes
Information Element (IEI) 1
Information Element Data Length (IEDL) 1
Information Element Data (IED) variable (IEDL defines length)
Table 2: The User Data Header (UDH).
Data Length and the DELIVER flags will be set to corresponding values in
order to create valid SMS DELIVER messages.
2.1.2 The User Data Header
The User Data Header (UDH) provides the means to add control information
to an SMS message in addition to the actual message payload or text. The
existence of a User Data Header is indicated through the User Data Header
Indication (UDHI) flag in the DELIVER field of an SMS DELIVER message.
If the flag is set, the header is present in the User Data of the message. The
User Data Header consists of the User Data Header Length (UDHL), followed
by one or multiple headers. The UDHL is the first byte of the User Data in
the SMS DELIVER message. The format for a single User Data Header is
shown in Table 2.
3 Mobile Phone Side SMS Delivery
Most current smart phones are build out of two processors. The main CPU,
called the application processor, is the processor that executes the smart
phone operating system and the user applications such as the mobile telephony
and the PIM applications. The second CPU runs a specialized real
time operating system that controls the mobile phone interface and is called
the modem. The modem handles all communication with the mobile phone
network and provides a control interface to the application processor.
Logically the application processor and the modem communicate through
one or multiple serial lines. The mobile telephony software stack running
on the application processor and communicates with the modem through a
text-command-based interface using a serial line interface provided by the
operating system running on the application processor. The physical connection
between the application processor and the modem solely depends
on the busses and interfaces offered by both sides but is irrelevant for our
The modems of our test devices (the iPhone, the HTC G1 Android, and
the HTC-Touch 3G Windows Mobile) are controlled through the GSM AT
command set. The GSM AT commands are used to control every aspect of
the mobile phone network interface, from network registration, call control
and SMS delivery to packet-based data connectivity.
3.1 The Telephony Stack
The telephony stack is the software component that handles all aspects of
the communication between the application processor and the modem. The
lowest layer in a telephony stack usually is a multiplexing layer to allow multiple
applications to access the modem at the same time. The multiplexing
layer also is the instance that translates API-calls to AT commands and AT
result codes to status messages. The applications to allow the user to place
and answer phone calls and to read and write short messages exist on top of
the multiplexing layer.
3.2 SMS Delivery
Short messages are delivered through unsolicited AT command result codes
issued by the modem to the application processor. The result code consists
+CMT: ,22
Figure 1: Unsolicited AT result code that indicates the reception of an SMS
of two lines of ASCII text. The first line contains the result code and the
number of bytes that follow on the second line. The number of bytes is given
as the number of octets after the hexadecimal to binary conversion. The
second line contains the entire SMS message in hexadecimal representation.
Figure 3 shows an example of an incoming SMS message using the CMT
result code which is used for SMS delivery on all of our test devices. There
are multiple result codes for delivering incoming SMS messages to the application
processor but our investigation results show that all our devices are
configured to use the CMT result code. Upon reception of the message the
application processor usually has to acknowledge the reception by issuing a
specific AT command to the modem. All interaction to the point of acknowledging
the reception of the CMT result is handled by the multiplexing layer
of the telephony stack.
3.3 The Stacks of our Test Devices
We will shortly describe the parts of the telephony stack that are relevant
for SMS handling on each of our test platforms.
3.3.1 iPhone OS
On the iPhone, the telephony stack mainly consists of one application binary
called CommCenter. CommCenter communicates directly with the modem
using a number of serial lines of which two are used for AT commands related
to SMS transfers. It handles incoming SMS messages by itself without
invoking any other process, besides when the device notifies the user about
a newly arrived message after storing it in the SMS database. The user SMS
application is only used for reading SMS messages stored in the database and
for composing new messages and does not itself directly communicate with
the modem.
3.3.2 Android
On the Android platform the telephony stack consists of the radio interface
layer (RIL) that takes the role of the multiplexing layer described above.
The RIL is a single daemon running on the device and communicates with
the modem through a single serial line. On top of the RIL daemon, the
Android phone application (com.android.phone) handles the communication
with the mobile phone network. The phone application receives incoming
SMS messages and forwards them to the SMS and MMS application
3.3.3 Windows Mobile
In Windows Mobile, the telephony stack is quite a bit larger and more distributed
compared with the iPhone and the Android telephony stacks. The
parts relevant to SMS are: the SmsRouter library (Sms Providers.dll) and
the tmail.exe binary. The tmail.exe binary is the SMS and MMS application
that provides a user interface for reading and composing SMS messages.
Other components such as theWAP PushRouter sit on top of the SmsRouter.
4 SMS Injection
Based on the results of our analysis on how SMS messages are delivered to
the application layer, we designed our SMS injection framework.
Our method for SMS injection is based on adding a layer between the
serial lines and the multiplexer (the lowest layer of the telephony stack). We
call this new layer the injector. The purpose of the injector is to perform a
man-in-the-middle attack on the communication between the modem and the
telephony stack. The basic functionality of the injector is to read commands
from the multiplexer and forward them to the modem and in return read
back the results from the modem and forward them to the multiplexer.
To inject an SMS message into the application layer, the injector generates
a new CMT result and sends it to the multiplexer just as it would forward a
real SMS message from the modem. It further handles the acknowledgement
commands sent by the multiplexer. Figure 2 shows the logical model of our
injection framework.
We implemented our injection framework for our three test platforms:
iPhone OS, Android, and Windows Mobile. We believe that our approach
Figure 2: Logical model of our injector framework.
for message injection can be easily ported to other smart phone platforms if
these allow application level access to the serial lines of the modem or the
ability to replace or add an additional driver that provides the serial line
We noticed several positive side effects of our framework, some of which
can be used to further improve the analysis process. First of all, we can
monitor and log all SMS messages being sent and received. This ability can
be used to analyze proprietary protocols based on SMS, such as the iPhone’s
visual voice mail. The ability to monitor all AT commands and responses
between the telephony stack and the modem provides an additional source
of feedback while conducting various tests. On the iPhone, for example,
messages are not acknowledged in a proper way if these contain unsupported
4.1 The Injection Framework
Below we will briefly describe the implementation issues of the injection
framework for each of our target platforms. Every implementation of the
framework opens TCP port 4223 on all network interfaces in order to receive
the SMS messages that should be injected. This network based approach
gives us a high degree of flexibility for implementing our testing tools independent
from the tested platform.
So far we are able to install our injection framework on all the test targets
and continue to use them as if the injection framework was not installed,
therefore giving us high degree of confidence in our approach.
4.2 iPhone
On the iPhone, SMS messages are handled by the CommCenter process.
CommCenter is the central control for all communication functionalities of
the iPhone such asWiFi, Bluetooth, and the 2G or 3G modem for voice, short
messages, and packeted-based communication. The interface for CommCenter
consists of sixteen virtual serial lines, /dev/dlci.h5-baseband.[0-15]
and /dev/dlci.spi-baseband.[0-15] on the 2G and the 3G iPhone, respectively.
The implementation of our injection framework for the iPhone OS is
separated into two parts, a library and a daemon. The library is injected into
the CommCenter process through library pre-loading. The library intercepts
the open(2) function from the standard C library. Our version of open checks
for access to the two serial lines used for AT commands. If the respective
files are opened the library replaces the file descriptor with one connected to
our daemon. The corresponding device files are the serial lines 3 and 4 on
the 2G and 3G iPhones. The library’s only function is to redirect the serial
lines to the daemon. The daemon implements the actual message injection
and log functionality.
Figure 3 shows the shell script that loads our injector framework on the
iPhone. In the first step the injector daemon is started. In the second step,
the library is copied in to the Libraries directory. Third, the plist responsible
for CommCenter is overwritten with our version that contains the necessary
settings in order to load our library that hijacks the open(2) call. Our plist
is shown in Figure 4.
4.3 Android
The implementation for the Android platform consists of just a single daemon.
The daemon talks directly to the serial line device connected to the
modem and emulates a new serial device through creation of a virtual terminal.
The injection framework is installed in three steps. First, the actual
serial line device is renamed from /dev/smd0 to /dev/smd0real. Second,
the daemon is started, opens /dev/smd0real and creates the emulated serial
device by creating a TTY named /dev/smd0. In the third step, the RIL
process (/system/bin/rild), is restarted by sending it the TERM signal.
Upon restart, rild opens the emulated serial line and from there on will talk
./injector &
mkdir -p /System/Library/Test/
rm -f /System/Library/Test/libopen*
cp libopen* /System/Library/Test/
cp com.apple.CommCenter.plist /System/Library/LaunchDaemons/
launchctl unload -w /System/Library/LaunchDaemons/
launchctl load -w /System/Library/LaunchDaemons/
Figure 3: Installing and starting the injector.



Figure 4: The plist entry that injects our library into CommCenter.
mv /dev/smd0 /dev/smd0real
/data/myrild &
kill 33 # PID of rild (/system/bin/rild)
Figure 5: Installing and starting the injector.
to our daemon instead of the modem.
Figure 5 shows the shell script the loads our injector on the Android
4.4 Windows Mobile
The Windows Mobile version of our injection framework is based on the simple
log-driver written by Willem Hengeveld. The original log-driver [4] was
designed for logging all AT communication between the user space process
and the modem. We added the injection and state tracking functionality. To
do this, we had to modify the driver quite a bit in order to have it listen
on the TCP port to connect our test tools. The driver replaces the original
serial driver and provides the same interface the original driver had and
loads the original driver in order to communicate with the modem. The
driver is installed through modifying several keys of the Windows Mobile
registry at: /HKEY LOCAL MACHINE/Drivers/BuiltIn/SMD0. The most important
change is the name of the Dynamic Link Library (DLL) that provides
the driver for the interface, whose key is named Dll. Its original value
is smd com.dll.
In order to install our injector DLL some steps need to be performed.
First the device has to be Application Unlocked. This is necessary in
order to install unsigned libraries. The unlocking can be performed through
changing certain registry values. The details of the Application Unlocking [1]
can be easily found on the Internet. Second, the installation steps from the
original logdev [4] driver need to be followed.
Once our injector framework is installed it can be activated and deactivated
through changing the value of the Dll key in /HKEY LOCAL MACHINE/
Drivers/BuiltIn/SMD0. Activate it by changing it from smd com.dll to
injector.dll. To deactivate change it back to smd com.dll. After the
change the device needs to be rebooted.
5 Fuzzing
Fuzzing is one of the easiest and most efficient ways to find implementation
vulnerabilities. With this framework, we are able to quickly inject fuzzed
SMS messages into the telephony stack by sending them over the listening
TCP port. In general, there are three basic steps in fuzzing. The first is test
generation. The second is delivering the test cases to the application, and
the final step is application monitoring. All of these steps are important to
find vulnerabilities with fuzzing.
5.0.1 Fuzzing Test Cases
We took a couple of approaches to generating the fuzzed SMS messages. One
was to write our own Python library which generated the test cases while
the other was to use the Sulley [2] fuzzing framework. In either case, the
most important part was to express a large number of different types of SMS
messages. Below are some examples of the types of messages that we fuzzed.
5.0.2 Basic SMS Messages
As from Table 1, we fuzzed various fields in a standard SMS message including
elements such as the sender address, the user data (or message), and the
various flags.
5.0.3 Basic UDH Messages
As seen in Table 2, we fuzzed various fields in the UDH header. This included
the UDH information element and UDH data.
5.0.4 Concatenated SMS Messages
Concatenation provides the means to compose SMS messages that exceed
the 140 byte (160 7-bit character) limitation. Concatenation is achieved
through the User Data Header type 0 as specified in [3]. The concatenation
header consist of five bytes, the type (IEI), the length (IEDL), and three
bytes of header data (IED) as seen in Table 3. By fuzzing these fields we
force messages to arrive out of order or not at all, as well as sending large
IED Byte Index Purpose
0 ID (same for all chunks)
1 Number of Chunks
2 Chunk Index
Table 3: The UDH for SMS Concatenation.
IED Byte Index Purpose
0 - 1 Destination Port (16bit)
2 - 3 Source Port (16bit)
Table 4: The UDH for SMS Port Addressing.
5.0.5 UDH Port Scanning
SMS applications can register to receive data on UDH ports, analogous to
the way TCP and UDP applications can do so. Without reverse engineering,
it is impossible to know exactly what ports a particular mobile OS will have
applications listening on. We send large amounts of (unformatted) data to
each port. The structure of the UDH destined for particular applications of
designated ports is indicated in Table 4.
5.0.6 Visual voice mail (iPhone only)
When a visual voice mail arrives, an SMS message arrives on port 5499 that
contains a URL in which the device can receive the actual voice mail audio
file. This URL is only accessible on the interface that connects to the AT&T
network, and will not connect to a generic URL on the Internet. The URL
is clearly to a web application that has variables encoded in the URL. We
fuzz the format of this URL.
5.1 Delivery
Once the test cases are generated, they need to be delivered to the appropriate
application. In this case, due to the way we have designed the testing
framework, it is possible to simply send them to a listening TCP port. All
of this work is designed to make it easy to deliver the test cases.
5.2 Monitoring
It does no good to generate and send fuzzed test cases if you do not know
when a problem occurs. Device monitoring is just as important as the other
steps. Unfortunately, monitoring is device dependent. There are two important
things to monitor. We need to know if a test case causes a crash. We
also need to know if a test case causes a degradation of service, i.e. if the
process does not crash but otherwise stops functioning properly.
5.2.1 iPhone
On the iPhone OS, the crash of a process causes a crash dump file to be
written to the file system compliments of Crash Reporter. This crash dump
can be retrieved and analyzed to determine the kind and position of the
crash. In between each fuzzed test case, a known valid test case is sent. The
SMS database can be queried to ensure that this test case was received and
recorded. If not, an error can be reported. In this fashion, it is possible to
detect errors that do not necessarily result in a crash.
5.2.2 Android
The Android development kit takes a different approach by suppling a tool
called the Android Debug Bridge (ADB), this tool allows us to monitor the
system log of the Android platform. If an application crashes on Android
the system log will contain the required information about the crash. If a
Java/Dalvik process crashes, it will contain information including the back
trace of the application. Like on the iPhone we query the SMS database to
be more certain to catch messages that cause problems but not necessarily
cause a crash.
5.2.3 Windows Mobile
The Windows Mobile development kit on the other hand provides the tools
for on-device debugging. This means Windows Mobile allows traditional
fuzzing by attaching a debugger to the process being fuzzed.
6 Results
We found multiple problematic SMS messages on the iPhone and on Android.
At the point of writing this white paper we are still busy fuzzing Windows
In order to determine if a problematic SMS message that was found using
our local injection framework can be abused for an attack it needs to be
determined if the specific message can be sent over the real mobile phone network
of an operator. The test is quite simple since the message body just has
to be copied from the generated SMS DILVER message to a SMS SUBMIT
message. The SMS SUBMIT message then can be easily sent using any mobile
phone or GSM/3G modem that supports sending of binary messages via
AT commands.
In the following we will briefly describe the bugs and vulnerabilities we
have found. Of course we will not reveil the actual SMS messages that cause
the crash until our presentation.
6.1 iPhone
Our iPhone OS targets were running OS version 2.2 and 2.2.1. We discovered
a couple of bugs of which we will describe two particularly interesting ones.
The first bug, or first kind of bug since we discovered multiple instances
of it, crashes SpringBoard (the iPhone OS window manager). When Spring-
Board crashes it interrupts any currently running application, when it restarts
it locks the device and forces the user to slide an unlock. If a pass code is
set the user has to enter the pass code. Crashing SpringBoard takes a couple
of seconds since it has to write a crash dump, therefore this bug can be
utilized for a Denial-of-Service attack where the victim is barred from using
his iPhone.
The second bug is much more interesting since it is able to crash the
CommCenter process. If you remember our description from earlier section
you know that CommCenter manages the iPhones connectivity. If Comm-
Center crashes the phone completely looses network connectivity (GSM/3G
and WiFi). If a phone call is in progress the call is interrupted. This bug
can be utilized for a serious Denial-of-Service attack since the victim can be
effectively barred from making and receiving phone calls. Figure 6 below
shows the result of such an SMS message, the iPhone searches for its mobile
phone network.
6.2 Android
Our Android targets were running OS version 1.0, 1.1, and 1.5. We found
several bugs on all Android versions of which we will discuss one particular
bug that is present in a different form on every Android version we tested.
The bug is similar to the second iPhone bug in the way that it kills the
telephony process (com.android.phone) and thus kicks the Android device
from the mobile phone network. On Android the bug is a little more interesting
since it will permanently kick the target device off the network if the
SIM card residing in the phone has a PIN set. The problem is that when
com.android.phone is restarted it resets the modem and therefore clears the
PIN. After the attack the Android phone is disconnected from the network
until the user enters the PIN of the SIM. Until the PIN is entered the user
cannot be called, does not receive SMS messages, and of course email and
other IP-based services are disabled. The attack is silent, the user is not
notified about the bad SMS in any way, and, therefore, very effective. The
only way to detect it is to regularly check the phone to see if it shows that the
SIM card is locked. Figure 7 shows the crashed com.android.phone process.
Figure 8 shows the Java/Dalvik trace of the crash and Figure 9 shows the
locked SIM card.
Figure 6: iPhone lost its network connectivity and starts searching.
Figure 7: com.android.phone crashed.
D/WAP PUSH( 7085): Rx: xxxxxxxx(blocked)xxxxxxxxxxx
W/dalvikvm( 7085): threadid=3: thread exiting with uncaught
exception (group=0x4000fe70)
E/AndroidRuntime( 7085): Uncaught handler: thread main exiting
due to uncaught exception
E/AndroidRuntime( 7085): java.lang.ArrayIndexOutOfBoundsException
E/AndroidRuntime( 7085): at com.android.internal.telephony.
E/AndroidRuntime( 7085): at com.android.internal.telephony.
E/AndroidRuntime( 7085): at com.android.internal.telephony.
E/AndroidRuntime( 7085): at com.android.internal.telephony.
E/AndroidRuntime( 7085): at com.android.internal.telephony.gsm.
E/AndroidRuntime( 7085): at com.android.internal.telephony.gsm.
E/AndroidRuntime( 7085): at android.os.Handler.dispatchMessage(
E/AndroidRuntime( 7085): at android.os.Looper.loop(Looper.java:123)
E/AndroidRuntime( 7085): at android.app.ActivityThread.main(
E/AndroidRuntime( 7085): at java.lang.reflect.Method.invokeNative(
Native Method)
E/AndroidRuntime( 7085): at java.lang.reflect.Method.invoke(
E/AndroidRuntime( 7085): at com.android.internal.os.ZygoteInit
E/AndroidRuntime( 7085): at com.android.internal.os.ZygoteInit.main(
E/AndroidRuntime( 7085): at dalvik.system.NativeStart.main(Native
I/Process ( 56): Sending signal. PID: 7085 SIG: 3
I/dalvikvm( 7085): threadid=7: reacting to signal 3
W/ActivityManager( 56): Process com.android.phone has crashed too
many times: killing!
Figure 8: Crash trace for com.android.phone.
Figure 9: The SIM card locked, phone is not connected to the network.
[1] http://robertpeloschek.blogspot.com/2006/03/
[2] Sulley - Pure Python fully automated and unattended fuzzing framework.
[3] 3rd Generation Partnership Project. 3GPP TS 23.040 - Technical realization
of the Short Message Service (SMS). http://www.3gpp.org/ftp/
Specs/html-info/23040.htm, September 2004.
[4] W. J. Hengeveld. Windows Mobile AT-command log-driver. http://