LibSEAL is a SEcure Auditing Library for internet services. It allows to detect integrity violations of internet services without the need to trust the service operator. To do so, LibSEAL: (i) constructs a secure log of requests and responses exchanged between clients and the internet service; and (ii) periodically checks the log for integrity violations. Technically, LibSEAL combines TaLoS, a secure TLS communication library, with SQLite in order to create and check logs. Further details about the operation of LibSEAL are described in the corresponding EuroSys'18 publication.
This repository contains the source code of LibSEAL. It ships as a set of patches on top of TaLoS. In particular, the repository ships with auditing modules for the Dropbox, Git and ownCloud services.
LibSEAL builds on top of the TaLoS library. You thus need to clone both the TaLoS and the LibSEAL repositories:
$ git clone https://github.com/lsds/TaLoS
$ git clone https://github.com/lsds/LibSEALWe assume that you have cloned TaLoS into the ${TALOS_ROOT} directory
(e.g. /home/<username>/talos/) and that you have cloned LibSEAL into
the ${LIBSEAL_ROOT} directory (eg /home/<username>/libseal/).
You first need to copy the LibSEAL patches into the TaLoS source directory:
$ cd ${TALOS_ROOT}/src
$ cp -r ${LIBSEAL_ROOT}/src/libseal .
$ cd talos && ./patch_libressl.sh && cd ..
$ cd libseal && ./patch_talos.sh && cd ..After that, please follow the instructions in the TaLoS readme file. Since you applied the LiBSEAL patches above, this will indeed compile LibSEAL.
By default, LibSEAL uses the Git auditing module. To use LiBSEAL with
modules (ownCloud or Dropbox), you need to change the
enclave.signed.so symlink and the auditing.so symlink to point to
the appropriate auditing module in directory
${TALOS_ROOT}/src/libressl-2.4.1/crypto.
LibSEAL ships with three auditing modules, for Dropbox, Git and ownCloud.
These are located in directory
${TALOS_ROOT}/src/libressl-2.4.1/crypto/auditing.
Each module defines:
- Code to process the service's HTTP requests/responses, extracting information to create the logs that will be checked for integrity violations;
- SQL queries to check for integrity violations. These queries return an empty set only if no violations have been detected. The definition of an integrity violation depends on the service. For example, for Dropbox, LibSEAL detects whether the list of files provided by Dropbox corresponds to the list of files that the client has uploaded;
- SQL queries to trim the log, i.e., to remove entries that are no longer needed to check the log. Trimming queries are used to reduce the log size.
LibSEAL can use a distributed monotonic counter service to prevent
rollback attacks. The implementation (files mcservice.c and
mcservice.h in directory
${LIBSEAL_ROOT}/src/monotoniccounterservice) is similar to the service
described in the ROTE paper.
To use this service, you first need to define the
USE_MONOTONIC_COUNTER_SERVICE macro in enclaveshim_config.h. Then,
you need to create a monotonic_counter_service.txt file that has to be
accessible by your application (see the mcservice_initialize()
function in mcservice.c). This file must contain the IP address of
every machine that will be used for the monotonic counter service. These
machines must start the server present in the
${LIBSEAL_ROOT}/src/monotoniccounterservice directory. Two scripts, to
start and stop a server, are provided in this directory.
Upon startup, LibSEAL will connect to these servers. It will then exchange (encrypted) messages for each database insert in order to increment the monotonic counter.
LibSEAL adds several new macros in the enclaveshim_config.h file:
DO_LOGGING: define this macro if you want to use the logging module;LOG_FOR_SQUID: define this macro if you are using Squid. This is to avoid logging both the messages transmitted between the client and the proxy and the messages transmitted between the proxy and the server;USE_MONOTONIC_COUNTER_SERVICE: define this macro to use the distributed monotonic counter service. See the previous section for more details;SQLITE_DB_NAME: this macro defines the path and name of the SQLite database that contains the log or:memory:for an in-memory database;SQLITE_DO_INSERT: define this macro to execute the database insertions. This macro was used in our experimental evaluation to measure the cost of the processing of HTTP messages without any database operation;SQLITE_ASYNC_MODE: define this macro to write the database asynchronously to disk;SQLITE_ADD_HASH: define this macro to add a hash to every database entry, for integrity;SQLITE_DO_SIGN: define this macro to add a signature to every database entry, for integrity;SQLITE_DO_CHECK: define this macro to periodically check for invariant violations;SQLITE_CHECKAFTER: this macro controls the period, in terms of number of entries inserted into the database, at which the SQL queries for revealing invariant violations are executed;SQLITE_DO_TRIMMING: define this macro to execute the trimming query, in order to remove from the log entries that are no longer necessary.