diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-15 17:07:52 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-15 17:07:52 +0000 |
commit | f0f453c916e279980df981c1e1dee0d167dc124e (patch) | |
tree | d09973c9f173820ade2dc814467d3e57df8a042d /docker_test.sh | |
parent | Initial commit. (diff) | |
download | ssh-audit-upstream.tar.xz ssh-audit-upstream.zip |
Adding upstream version 3.1.0.upstream/3.1.0upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rwxr-xr-x | docker_test.sh | 801 |
1 files changed, 801 insertions, 0 deletions
diff --git a/docker_test.sh b/docker_test.sh new file mode 100755 index 0000000..99c1155 --- /dev/null +++ b/docker_test.sh @@ -0,0 +1,801 @@ +#!/bin/bash + +# +# This script will set up a docker image with multiple versions of OpenSSH, then +# use it to run tests. +# +# Optional arguments: +# --accept: accepts test failures and overwrites expected results with actual results (useful for updating the tests themselves). +# --create: attempts to create a new docker image. +# +# +# For debugging purposes, here is a cheat sheet for manually running the docker image: +# +# docker run -p 2222:22 -it ssh-audit-test:X /bin/bash +# docker run -p 2222:22 --security-opt seccomp:unconfined -it ssh-audit-test /debug.sh +# docker run -d -p 2222:22 ssh-audit-test:X /openssh/sshd-5.6p1 -D -f /etc/ssh/sshd_config-5.6p1_test1 +# docker run -d -p 2222:22 ssh-audit-test:X /openssh/sshd-8.0p1 -D -f /etc/ssh/sshd_config-8.0p1_test1 +# + + +# This is the docker tag for the image. If this tag doesn't exist, then we assume the +# image is out of date, and generate a new one with this tag. +IMAGE_VERSION=3 + +# This is the name of our docker image. +IMAGE_NAME=positronsecurity/ssh-audit-test-framework + + +# Terminal colors. +CLR="\033[0m" +#RED="\033[0;31m" +#YELLOW="\033[0;33m" +GREEN="\033[0;32m" +REDB="\033[1;31m" # Red + bold +YELLOWB="\033[1;33m" # Yellow + bold +GREENB="\033[1;32m" # Green + bold + +# Program return values. +PROGRAM_RETVAL_FAILURE=3 +PROGRAM_RETVAL_WARNING=2 +#PROGRAM_RETVAL_CONNECTION_ERROR=1 +PROGRAM_RETVAL_GOOD=0 + + +# Counts the number of test failures. +num_failures=0 + +# When set, if a failure is encountered, overwrite the expected output with the actual value (i.e.: the user validated the failures already and wants to update the tests themselves). +accept=0 + + +# Returns 0 if current docker image exists. +check_if_docker_image_exists() { + images=$(docker image ls | grep -E "$IMAGE_NAME[[:space:]]+$IMAGE_VERSION") +} + + +# Uncompresses and compiles the specified version of Dropbear. +compile_dropbear() { + version=$1 + compile "Dropbear" "$version" +} + + +# Uncompresses and compiles the specified version of OpenSSH. +compile_openssh() { + version=$1 + compile "OpenSSH" "$version" +} + + +# Uncompresses and compiles the specified version of TinySSH. +compile_tinyssh() { + version=$1 + compile "TinySSH" "$version" +} + + +compile() { + project=$1 + version=$2 + + tarball= + uncompress_options= + source_dir= + server_executable= + if [[ $project == "OpenSSH" ]]; then + tarball="openssh-${version}.tar.gz" + uncompress_options="xzf" + source_dir="openssh-${version}" + server_executable=sshd + elif [[ $project == "Dropbear" ]]; then + tarball="dropbear-${version}.tar.bz2" + uncompress_options="xjf" + source_dir="dropbear-${version}" + server_executable=dropbear + elif [[ $project == "TinySSH" ]]; then + tarball="${version}.tar.gz" + uncompress_options="xzf" + source_dir="tinyssh-${version}" + server_executable="build/bin/tinysshd" + fi + + echo "Uncompressing ${project} ${version}..." + tar $uncompress_options "$tarball" + + echo "Compiling ${project} ${version}..." + pushd "$source_dir" || exit > /dev/null + + # TinySSH has no configure script... only a Makefile. + if [[ $project == "TinySSH" ]]; then + make -j 10 + else + ./configure && make -j 10 + fi + + if [[ ! -f $server_executable ]]; then + echo -e "${REDB}Error: ${server_executable} not built!${CLR}" + exit 1 + fi + + echo -e "\n${GREEN}Successfully built ${project} ${version}${CLR}\n" + popd || exit > /dev/null +} + + +# Creates a new docker image. +create_docker_image() { + # Create a new temporary directory. + TMP_DIR=$(mktemp -d /tmp/sshaudit-docker-XXXXXXXXXX) + + # Copy the Dockerfile and all files in the test/docker/ dir to our new temp directory. + find test/docker/ -maxdepth 1 -type f -exec cp -t "$TMP_DIR" "{}" + + + # Make the temp directory our working directory for the duration of the build + # process. + pushd "$TMP_DIR" || exit > /dev/null + + # Get the release keys. + get_dropbear_release_key + get_openssh_release_key + get_tinyssh_release_key + + # Aside from checking the GPG signatures, we also compare against this known-good + # SHA-256 hash just in case. + get_openssh "4.0p1" "5adb9b2c2002650e15216bf94ed9db9541d9a17c96fcd876784861a8890bc92b" + get_openssh "5.6p1" "538af53b2b8162c21a293bb004ae2bdb141abd250f61b4cea55244749f3c6c2b" + get_openssh "8.0p1" "bd943879e69498e8031eb6b7f44d08cdc37d59a7ab689aa0b437320c3481fd68" + get_dropbear "2019.78" "525965971272270995364a0eb01f35180d793182e63dd0b0c3eb0292291644a4" + get_tinyssh "20190101" "554a9a94e53b370f0cd0c5fbbd322c34d1f695cbcea6a6a32dcb8c9f595b3fea" + + # Compile the versions of OpenSSH. + compile_openssh "4.0p1" + compile_openssh "5.6p1" + compile_openssh "8.0p1" + + # Compile the versions of Dropbear. + compile_dropbear "2019.78" + + # Compile the versions of TinySSH. + compile_tinyssh "20190101" + + + # Rename the default config files so we know they are our originals. + mv openssh-4.0p1/sshd_config sshd_config-4.0p1_orig + mv openssh-5.6p1/sshd_config sshd_config-5.6p1_orig + mv openssh-8.0p1/sshd_config sshd_config-8.0p1_orig + + + # Create the configurations for each test. + + + # + # OpenSSH v4.0p1 + # + + # Test 1: Basic test. + create_openssh_config "4.0p1" "test1" "HostKey /etc/ssh/ssh1_host_key\nHostKey /etc/ssh/ssh_host_rsa_key_1024\nHostKey /etc/ssh/ssh_host_dsa_key" + + + # + # OpenSSH v5.6p1 + # + + # Test 1: Basic test. + create_openssh_config "5.6p1" "test1" "HostKey /etc/ssh/ssh_host_rsa_key_1024\nHostKey /etc/ssh/ssh_host_dsa_key" + + # Test 2: RSA 1024 host key with RSA 1024 certificate. + create_openssh_config "5.6p1" "test2" "HostKey /etc/ssh/ssh_host_rsa_key_1024\nHostCertificate /etc/ssh/ssh_host_rsa_key_1024-cert_1024.pub" + + # Test 3: RSA 1024 host key with RSA 3072 certificate. + create_openssh_config "5.6p1" "test3" "HostKey /etc/ssh/ssh_host_rsa_key_1024\nHostCertificate /etc/ssh/ssh_host_rsa_key_1024-cert_3072.pub" + + # Test 4: RSA 3072 host key with RSA 1024 certificate. + create_openssh_config "5.6p1" "test4" "HostKey /etc/ssh/ssh_host_rsa_key_3072\nHostCertificate /etc/ssh/ssh_host_rsa_key_3072-cert_1024.pub" + + # Test 5: RSA 3072 host key with RSA 3072 certificate. + create_openssh_config "5.6p1" "test5" "HostKey /etc/ssh/ssh_host_rsa_key_3072\nHostCertificate /etc/ssh/ssh_host_rsa_key_3072-cert_3072.pub" + + + # + # OpenSSH v8.0p1 + # + + # Test 1: Basic test. + create_openssh_config "8.0p" "test1" "HostKey /etc/ssh/ssh_host_rsa_key_3072\nHostKey /etc/ssh/ssh_host_ecdsa_key\nHostKey /etc/ssh/ssh_host_ed25519_key" + + # Test 2: ED25519 certificate test. + create_openssh_config "8.0p1" "test2" "HostKey /etc/ssh/ssh_host_ed25519_key\nHostCertificate /etc/ssh/ssh_host_ed25519_key-cert.pub" + + # Test 3: Hardened installation test. + create_openssh_config "8.0p1" "test3" "HostKey /etc/ssh/ssh_host_ed25519_key\nKexAlgorithms curve25519-sha256,curve25519-sha256@libssh.org,diffie-hellman-group-exchange-sha256\nCiphers chacha20-poly1305@openssh.com,aes256-gcm@openssh.com,aes128-gcm@openssh.com,aes256-ctr,aes192-ctr,aes128-ctr\nMACs hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,umac-128-etm@openssh.com" + + + # Now build the docker image! + docker build --tag "$IMAGE_NAME:$IMAGE_VERSION" . + + popd || exit > /dev/null + rm -rf -- "$TMP_DIR" +} + + +# Creates an OpenSSH configuration file for a specific test. +create_openssh_config() { + openssh_version=$1 + test_number=$2 + config_text=$3 + + cp "sshd_config-${openssh_version}_orig" "sshd_config-${openssh_version}_${test_number}" + echo -e "${config_text}" >> "sshd_config-${openssh_version}_${test_number}" +} + + +# Downloads the Dropbear release key and adds it to the local keyring. +get_dropbear_release_key() { + get_release_key "Dropbear" "https://matt.ucc.asn.au/dropbear/releases/dropbear-key-2015.asc" "F29C6773" "F734 7EF2 EE2E 07A2 6762 8CA9 4493 1494 F29C 6773" +} + + +# Downloads the OpenSSH release key and adds it to the local keyring. +get_openssh_release_key() { + get_release_key "OpenSSH" "https://ftp.openbsd.org/pub/OpenBSD/OpenSSH/RELEASE_KEY.asc" "6D920D30" "59C2 118E D206 D927 E667 EBE3 D3E5 F56B 6D92 0D30" +} + + +# Downloads the TinySSH release key and adds it to the local keyring. +get_tinyssh_release_key() { + get_release_key "TinySSH" "" "96939FF9" "AADF 2EDF 5529 F170 2772 C8A2 DEC4 D246 931E F49B" +} + + +get_release_key() { + project=$1 + key_url=$2 + key_id=$3 + release_key_fingerprint_expected=$4 + + # The TinySSH release key isn't on any website, apparently. + if [[ $project == "TinySSH" ]]; then + gpg --keyserver keys.gnupg.net --recv-key "$key_id" + else + echo -e "\nGetting ${project} release key...\n" + wget -O key.asc "$2" + + echo -e "\nImporting ${project} release key...\n" + gpg --import key.asc + + rm key.asc + fi + + local release_key_fingerprint_actual + release_key_fingerprint_actual=$(gpg --fingerprint "$key_id") + if [[ $release_key_fingerprint_actual != *"$release_key_fingerprint_expected"* ]]; then + echo -e "\n${REDB}Error: ${project} release key fingerprint does not match expected value!\n\tExpected: $release_key_fingerprint_expected\n\tActual: $release_key_fingerprint_actual\n\nTerminating.${CLR}" + exit 1 + fi + echo -e "\n\n${GREEN}${project} release key matches expected value.${CLR}\n" +} + + +# Downloads the specified version of Dropbear. +get_dropbear() { + version=$1 + tarball_checksum_expected=$2 + get_source "Dropbear" "$version" "$tarball_checksum_expected" +} + + +# Downloads the specified version of OpenSSH. +get_openssh() { + version=$1 + tarball_checksum_expected=$2 + get_source "OpenSSH" "$version" "$tarball_checksum_expected" +} + + +# Downloads the specified version of TinySSH. +get_tinyssh() { + version=$1 + tarball_checksum_expected=$2 + get_source "TinySSH" "$version" "$tarball_checksum_expected" +} + + +get_source() { + project=$1 + version=$2 + tarball_checksum_expected=$3 + + base_url_source= + base_url_sig= + tarball= + sig= + signer= + if [[ $project == "OpenSSH" ]]; then + base_url_source="https://cdn.openbsd.org/pub/OpenBSD/OpenSSH/portable/" + base_url_sig=$base_url_source + tarball="openssh-${version}.tar.gz" + sig="${tarball}.asc" + signer="Damien Miller " + elif [[ $project == "Dropbear" ]]; then + base_url_source="https://matt.ucc.asn.au/dropbear/releases/" + base_url_sig=$base_url_source + tarball="dropbear-${version}.tar.bz2" + sig="${tarball}.asc" + signer="Dropbear SSH Release Signing <matt@ucc.asn.au>" + elif [[ $project == "TinySSH" ]]; then + base_url_source="https://github.com/janmojzis/tinyssh/archive/" + base_url_sig="https://github.com/janmojzis/tinyssh/releases/download/${version}/" + tarball="${version}.tar.gz" + sig="${tarball}.asc" + signer="Jan Mojžíš <jan.mojzis@gmail.com>" + fi + + echo -e "\nGetting ${project} ${version} sources...\n" + wget "${base_url_source}${tarball}" + + echo -e "\nGetting ${project} ${version} signature...\n" + wget "${base_url_sig}${sig}" + + + # Older OpenSSH releases were .sigs. + if [[ ($project == "OpenSSH") && (! -f $sig) ]]; then + wget "${base_url_sig}openssh-${version}.tar.gz.sig" + sig=openssh-${version}.tar.gz.sig + fi + + local gpg_verify + gpg_verify=$(gpg --verify "${sig}" "${tarball}" 2>&1) + retval=$? + if [[ $gpg_verify != *"Good signature from \"${signer}"* ]]; then + echo -e "\n\n${REDB}Error: ${project} signature invalid!\n$gpg_verify\n\nTerminating.${CLR}" + exit 1 + fi + + # Check GPG's return value. 0 denotes a valid signature, and 1 is returned + # on invalid signatures. + if [[ ${retval} != 0 ]]; then + echo -e "\n\n${REDB}Error: ${project} signature invalid! Verification returned code: $?\n\nTerminating.${CLR}" + exit 1 + fi + + echo -e "${GREEN}Signature on ${project} sources verified.${CLR}\n" + + local checksum_actual + checksum_actual=$(sha256sum "${tarball}" | cut -f1 -d" ") + if [[ $checksum_actual != "$tarball_checksum_expected" ]]; then + echo -e "${REDB}Error: ${project} checksum is invalid!\n Expected: ${tarball_checksum_expected}\n Actual: ${checksum_actual}\n\n Terminating.${CLR}" + exit 1 + fi +} + + +# Pulls the defined image from Dockerhub. +pull_docker_image() { + docker pull "${IMAGE_NAME}:${IMAGE_VERSION}" + retval=$? + if [[ ${retval} == 0 ]]; then + echo -e "${GREEN}Successfully downloaded image ${IMAGE_NAME}:${IMAGE_VERSION} from Dockerhub.${CLR}\n" + else + echo -e "${REDB}Failed to pull image ${IMAGE_NAME}:${IMAGE_VERSION} from Dockerhub! Error code: ${retval}${CLR}\n" + exit 1 + fi +} + + +# Runs a Dropbear test. Upon failure, a diff between the expected and actual results +# is shown, then the script immediately terminates. +run_dropbear_test() { + dropbear_version=$1 + test_number=$2 + options=$3 + expected_retval=$4 + + run_test "Dropbear" "${dropbear_version}" "${test_number}" "${options}" "${expected_retval}" +} + + +# Runs an OpenSSH test. Upon failure, a diff between the expected and actual results +# is shown, then the script immediately terminates. +run_openssh_test() { + openssh_version=$1 + test_number=$2 + expected_retval=$3 + + run_test "OpenSSH" "${openssh_version}" "${test_number}" "" "${expected_retval}" +} + + +# Runs a TinySSH test. Upon failure, a diff between the expected and actual results +# is shown, then the script immediately terminates. +run_tinyssh_test() { + tinyssh_version=$1 + test_number=$2 + expected_retval=$3 + + run_test "TinySSH" "${tinyssh_version}" "${test_number}" "" "${expected_retval}" +} + + +run_test() { + server_type=$1 + version=$2 + test_number=$3 + options=$4 + expected_retval=$5 + + failed=0 # Set to 1 if this test fails. + server_exec= + test_result_stdout= + test_result_json= + expected_result_stdout= + expected_result_json= + test_name= + if [[ $server_type == "OpenSSH" ]]; then + server_exec="/openssh/sshd-${version} -D -f /etc/ssh/sshd_config-${version}_${test_number}" + test_result_stdout="${TEST_RESULT_DIR}/openssh_${version}_${test_number}.txt" + test_result_json="${TEST_RESULT_DIR}/openssh_${version}_${test_number}.json" + expected_result_stdout="test/docker/expected_results/openssh_${version}_${test_number}.txt" + expected_result_json="test/docker/expected_results/openssh_${version}_${test_number}.json" + test_name="OpenSSH ${version} ${test_number}" + options= + elif [[ $server_type == "Dropbear" ]]; then + server_exec="/dropbear/dropbear-${version} -F ${options}" + test_result_stdout="${TEST_RESULT_DIR}/dropbear_${version}_${test_number}.txt" + test_result_json="${TEST_RESULT_DIR}/dropbear_${version}_${test_number}.json" + expected_result_stdout="test/docker/expected_results/dropbear_${version}_${test_number}.txt" + expected_result_json="test/docker/expected_results/dropbear_${version}_${test_number}.json" + test_name="Dropbear ${version} ${test_number}" + elif [[ $server_type == "TinySSH" ]]; then + server_exec="/usr/bin/tcpserver -HRDl0 0.0.0.0 22 /tinysshd/tinyssh-20190101 -v /etc/tinyssh/" + test_result_stdout="${TEST_RESULT_DIR}/tinyssh_${version}_${test_number}.txt" + test_result_json="${TEST_RESULT_DIR}/tinyssh_${version}_${test_number}.json" + expected_result_stdout="test/docker/expected_results/tinyssh_${version}_${test_number}.txt" + expected_result_json="test/docker/expected_results/tinyssh_${version}_${test_number}.json" + test_name="TinySSH ${version} ${test_number}" + fi + + #echo "Running: docker run --rm -d -p 2222:22 $IMAGE_NAME:$IMAGE_VERSION ${server_exec}" + cid=$(docker run --rm -d -p 2222:22 "${IMAGE_NAME}:${IMAGE_VERSION}" ${server_exec}) + retval=$? + if [[ ${retval} != 0 ]]; then + echo -e "${REDB}Failed to run docker image! (exit code: ${retval})${CLR}" + exit 1 + fi + + ./ssh-audit.py localhost:2222 > "$test_result_stdout" + actual_retval=$? + if [[ $actual_retval != "$expected_retval" ]]; then + echo -e "${REDB}Unexpected return value. Expected: ${expected_retval}; Actual: ${actual_retval}${CLR}" + + if [[ $accept == 1 ]]; then + echo -e "\n${REDB}This failure cannot be automatically fixed; this script must be manually updated with the new expected return value.${CLR}" + fi + + cat "${test_result_stdout}" + docker container stop -t 0 "${cid}" > /dev/null + exit 1 + fi + + ./ssh-audit.py -jj localhost:2222 > "$test_result_json" + actual_retval=$? + if [[ $actual_retval != "$expected_retval" ]]; then + echo -e "${REDB}Unexpected return value. Expected: ${expected_retval}; Actual: ${actual_retval}${CLR}" + + if [[ $accept == 1 ]]; then + echo -e "\n${REDB}This failure cannot be automatically fixed; this script must be manually updated with the new expected return value.${CLR}" + fi + + cat "${test_result_json}" + docker container stop -t 0 "${cid}" > /dev/null + exit 1 + fi + + docker container stop -t 0 "${cid}" > /dev/null + retval=$? + if [[ ${retval} != 0 ]]; then + echo -e "${REDB}Failed to stop docker container ${cid}! (exit code: ${retval})${CLR}" + exit 1 + fi + + # TinySSH outputs a random string in each banner, which breaks our test. So + # we need to filter out the banner part of the output so we get stable, repeatable + # results. + if [[ $server_type == "TinySSH" ]]; then + grep -v "(gen) banner: " "${test_result_stdout}" > "${test_result_stdout}.tmp" + mv "${test_result_stdout}.tmp" "${test_result_stdout}" + cat "${test_result_json}" | perl -pe 's/"comments": ".*?"/"comments": ""/' | perl -pe 's/"raw": ".+?"/"raw": ""/' > "${test_result_json}.tmp" + mv "${test_result_json}.tmp" "${test_result_json}" + fi + + diff=$(diff -u "${expected_result_stdout}" "${test_result_stdout}") + retval=$? + if [[ ${retval} != 0 ]]; then + + # If the user wants to update the tests, then overwrite the expected results with the actual results. + if [[ $accept == 1 ]]; then + cp "${test_result_stdout}" "${expected_result_stdout}" + echo -e "${test_name} ${YELLOWB}UPDATED${CLR}\n" + else + echo -e "${test_name} ${REDB}FAILED${CLR}.\n\n${diff}\n" + failed=1 + num_failures=$((num_failures+1)) + fi + + fi + + diff=$(diff -u "${expected_result_json}" "${test_result_json}") + retval=$? + if [[ ${retval} != 0 ]]; then + + # If the user wants to update the tests, then overwrite the expected results with the actual results. + if [[ $accept == 1 ]]; then + cp "${test_result_json}" "${expected_result_json}" + echo -e "${test_name} ${YELLOWB}UPDATED${CLR}\n" + else + echo -e "${test_name} ${REDB}FAILED${CLR}.\n\n${diff}\n" + failed=1 + num_failures=$((num_failures+1)) + fi + + fi + + if [[ $failed == 0 ]]; then + echo -e "${test_name} ${GREEN}passed${CLR}." + fi +} + +run_builtin_policy_test() { + policy_name=$1 # The built-in policy name to use. + version=$2 # Version of OpenSSH to test with. + test_number=$3 # The test number to run. + server_options=$4 # The options to start the server with (i.e.: "-o option1,options2,...") + expected_exit_code=$5 # The expected exit code of ssh-audit.py. + + server_exec="/openssh/sshd-${version} -D -f /etc/ssh/sshd_config-8.0p1_test1 ${server_options}" + test_result_stdout="${TEST_RESULT_DIR}/openssh_${version}_builtin_policy_${test_number}.txt" + test_result_json="${TEST_RESULT_DIR}/openssh_${version}_builtin_policy_${test_number}.json" + expected_result_stdout="test/docker/expected_results/openssh_${version}_builtin_policy_${test_number}.txt" + expected_result_json="test/docker/expected_results/openssh_${version}_builtin_policy_${test_number}.json" + test_name="OpenSSH ${version} built-in policy ${test_number}" + + run_policy_test "${test_name}" "${server_exec}" "${policy_name}" "${test_result_stdout}" "${test_result_json}" "${expected_exit_code}" +} + + +run_custom_policy_test() { + config_number=$1 # The configuration number to use. + test_number=$2 # The policy test number to run. + expected_exit_code=$3 # The expected exit code of ssh-audit.py. + + version= + config= + if [[ ${config_number} == "config1" ]]; then + version="5.6p1" + config="sshd_config-5.6p1_test1" + elif [[ ${config_number} == "config2" ]]; then + version="8.0p1" + config="sshd_config-8.0p1_test1" + elif [[ ${config_number} == "config3" ]]; then + version="5.6p1" + config="sshd_config-5.6p1_test4" + fi + + server_exec="/openssh/sshd-${version} -D -f /etc/ssh/${config}" + policy_path="test/docker/policies/policy_${test_number}.txt" + test_result_stdout="${TEST_RESULT_DIR}/openssh_${version}_custom_policy_${test_number}.txt" + test_result_json="${TEST_RESULT_DIR}/openssh_${version}_custom_policy_${test_number}.json" + expected_result_stdout="test/docker/expected_results/openssh_${version}_custom_policy_${test_number}.txt" + expected_result_json="test/docker/expected_results/openssh_${version}_custom_policy_${test_number}.json" + test_name="OpenSSH ${version} custom policy ${test_number}" + + run_policy_test "${test_name}" "${server_exec}" "${policy_path}" "${test_result_stdout}" "${test_result_json}" "${expected_exit_code}" +} + + +run_policy_test() { + test_name=$1 + server_exec=$2 + policy_path=$3 + test_result_stdout=$4 + test_result_json=$5 + expected_exit_code=$6 + + + #echo "Running: docker run --rm -d -p 2222:22 $IMAGE_NAME:$IMAGE_VERSION ${server_exec}" + cid=$(docker run --rm -d -p 2222:22 "${IMAGE_NAME}:${IMAGE_VERSION}" ${server_exec}) + retval=$? + if [[ ${retval} != 0 ]]; then + echo -e "${REDB}Failed to run docker image! (exit code: ${retval})${CLR}" + exit 1 + fi + + #echo "Running: ./ssh-audit.py -P \"${policy_path}\" localhost:2222 > ${test_result_stdout}" + ./ssh-audit.py -P "${policy_path}" localhost:2222 > "${test_result_stdout}" + actual_exit_code=$? + if [[ ${actual_exit_code} != "${expected_exit_code}" ]]; then + echo -e "${test_name} ${REDB}FAILED${CLR} (expected exit code: ${expected_exit_code}; actual exit code: ${actual_exit_code}\n" + + if [[ $accept == 1 ]]; then + echo -e "\n${REDB}This failure cannot be automatically fixed; this script must be manually updated with the new expected return value.${CLR}" + fi + + cat "${test_result_stdout}" + docker container stop -t 0 "${cid}" > /dev/null + exit 1 + fi + + #echo "Running: ./ssh-audit.py -P \"${policy_path}\" -jj localhost:2222 > ${test_result_json} 2> /dev/null" + ./ssh-audit.py -P "${policy_path}" -jj localhost:2222 > "${test_result_json}" 2> /dev/null + actual_exit_code=$? + if [[ ${actual_exit_code} != "${expected_exit_code}" ]]; then + echo -e "${test_name} ${REDB}FAILED${CLR} (expected exit code: ${expected_exit_code}; actual exit code: ${actual_exit_code}\n" + + if [[ $accept == 1 ]]; then + echo -e "\n${REDB}This failure cannot be automatically fixed; this script must be manually updated with the new expected return value.${CLR}" + fi + + cat "${test_result_json}" + docker container stop -t 0 "${cid}" > /dev/null + exit 1 + fi + + docker container stop -t 0 "${cid}" > /dev/null + retval=$? + if [[ ${retval} != 0 ]]; then + echo -e "${REDB}Failed to stop docker container ${cid}! (exit code: ${retval})${CLR}" + exit 1 + fi + + diff=$(diff -u "${expected_result_stdout}" "${test_result_stdout}") + retval=$? + if [[ ${retval} != 0 ]]; then + + # If the user wants to update the tests, then overwrite the expected results with the actual results. + if [[ $accept == 1 ]]; then + cp "${test_result_stdout}" "${expected_result_stdout}" + echo -e "${test_name} ${YELLOWB}UPDATED${CLR}\n" + else + echo -e "${test_name} ${REDB}FAILED${CLR}.\n\n${diff}\n" + exit 1 + fi + + fi + + diff=$(diff -u "${expected_result_json}" "${test_result_json}") + retval=$? + if [[ ${retval} != 0 ]]; then + + # If the user wants to update the tests, then overwrite the expected results with the actual results. + if [[ $accept == 1 ]]; then + cp "${test_result_json}" "${expected_result_json}" + echo -e "${test_name} ${YELLOWB}UPDATED${CLR}\n" + else + echo -e "${test_name} ${REDB}FAILED${CLR}.\n\n${diff}\n" + exit 1 + fi + + fi + + echo -e "${test_name} ${GREEN}passed${CLR}." +} + + +# First check if docker is functional. +docker version > /dev/null +retval=$? +if [[ ${retval} != 0 ]]; then + echo -e "${REDB}Error: 'docker version' command failed (error code: ${retval}). Is docker installed and functioning?${CLR}" + exit 1 +fi + + +# Check if the docker image is the most up-to-date version. +docker_image_exists=0 +check_if_docker_image_exists +retval=$? +if [[ ${retval} == 0 ]]; then + docker_image_exists=1 +fi + + +# Check if the user specified --create to build a new image. +if [[ ($# == 1) && ($1 == "--create") ]]; then + # Ensure that the image name doesn't already exist before building. + if [[ ${docker_image_exists} == 1 ]]; then + echo -e "${REDB}Error: --create specified, but ${IMAGE_NAME}:${IMAGE_VERSION} already exists!${CLR}" + exit 1 + else + echo -e "\nCreating docker image ${IMAGE_NAME}:${IMAGE_VERSION}..." + create_docker_image + echo -e "\n${GREEN}Done creating docker image!${CLR}" + exit 0 + fi +fi + + +# If the user passes --accept, then the actual results will replace the expected results (meaning the user wants to update the tests themselves due to new functionality). +if [[ ($# == 1) && ($1 == "--accept") ]]; then + accept=1 + echo -e "\n${YELLOWB}Expected test results will be replaced with actual results.${CLR}" +fi + + +# If we weren't explicitly told to create a new image, and it doesn't exist, then pull it from Dockerhub. +if [[ ${docker_image_exists} == 0 ]]; then + echo -e "\nPulling docker image ${IMAGE_NAME}:${IMAGE_VERSION}..." + pull_docker_image +fi + + +echo -e "\n${GREEN}Starting tests...${CLR}" + +# Create a temporary directory to write test results to. +TEST_RESULT_DIR=$(mktemp -d /tmp/ssh-audit_test-results_XXXXXXXXXX) + +# Now run all the tests. +echo -e "\nRunning tests..." +run_openssh_test "4.0p1" "test1" "${PROGRAM_RETVAL_FAILURE}" +echo +run_openssh_test "5.6p1" "test1" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "5.6p1" "test2" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "5.6p1" "test3" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "5.6p1" "test4" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "5.6p1" "test5" "${PROGRAM_RETVAL_FAILURE}" +echo +run_openssh_test "8.0p1" "test1" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "8.0p1" "test2" "${PROGRAM_RETVAL_FAILURE}" +run_openssh_test "8.0p1" "test3" "${PROGRAM_RETVAL_WARNING}" +echo +run_dropbear_test "2019.78" "test1" "-r /etc/dropbear/dropbear_rsa_host_key_1024 -r /etc/dropbear/dropbear_dss_host_key -r /etc/dropbear/dropbear_ecdsa_host_key" 3 +echo +run_tinyssh_test "20190101" "test1" "${PROGRAM_RETVAL_WARNING}" +echo +echo +run_custom_policy_test "config1" "test1" "${PROGRAM_RETVAL_GOOD}" +run_custom_policy_test "config1" "test2" "${PROGRAM_RETVAL_FAILURE}" +run_custom_policy_test "config1" "test3" "${PROGRAM_RETVAL_FAILURE}" +run_custom_policy_test "config1" "test4" "${PROGRAM_RETVAL_FAILURE}" +run_custom_policy_test "config1" "test5" "${PROGRAM_RETVAL_FAILURE}" +run_custom_policy_test "config2" "test6" "${PROGRAM_RETVAL_GOOD}" + +# Passing test with host key certificate and CA key certificates. +run_custom_policy_test "config3" "test7" "${PROGRAM_RETVAL_GOOD}" + +# Failing test with host key certificate and non-compliant CA key length. +run_custom_policy_test "config3" "test8" "${PROGRAM_RETVAL_FAILURE}" + +# Failing test with non-compliant host key certificate and CA key certificate. +run_custom_policy_test "config3" "test9" "${PROGRAM_RETVAL_FAILURE}" + +# Failing test with non-compliant host key certificate and non-compliant CA key certificate. +run_custom_policy_test "config3" "test10" "${PROGRAM_RETVAL_FAILURE}" + +# Passing test with host key size check. +run_custom_policy_test "config2" "test11" "${PROGRAM_RETVAL_GOOD}" + +# Failing test with non-compliant host key size check. +run_custom_policy_test "config2" "test12" "${PROGRAM_RETVAL_FAILURE}" + +# Passing test with DH modulus test. +run_custom_policy_test "config2" "test13" "${PROGRAM_RETVAL_GOOD}" + +# Failing test with DH modulus test. +run_custom_policy_test "config2" "test14" "${PROGRAM_RETVAL_FAILURE}" + +# Failing test for built-in OpenSSH 8.0p1 server policy (RSA host key size is 3072 instead of 4096). +run_builtin_policy_test "Hardened OpenSSH Server v8.0 (version 4)" "8.0p1" "test1" "-o HostKeyAlgorithms=rsa-sha2-512,rsa-sha2-256,ssh-ed25519 -o KexAlgorithms=curve25519-sha256,curve25519-sha256@libssh.org,diffie-hellman-group16-sha512,diffie-hellman-group18-sha512,diffie-hellman-group-exchange-sha256 -o Ciphers=chacha20-poly1305@openssh.com,aes256-gcm@openssh.com,aes128-gcm@openssh.com,aes256-ctr,aes192-ctr,aes128-ctr -o MACs=hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,umac-128-etm@openssh.com" "${PROGRAM_RETVAL_FAILURE}" + +# Failing test for built-in OpenSSH 8.0p1 server policy (MACs not hardened). +run_builtin_policy_test "Hardened OpenSSH Server v8.0 (version 4)" "8.0p1" "test2" "-o HostKeyAlgorithms=rsa-sha2-512,rsa-sha2-256,ssh-ed25519 -o KexAlgorithms=curve25519-sha256,curve25519-sha256@libssh.org,diffie-hellman-group16-sha512,diffie-hellman-group18-sha512,diffie-hellman-group-exchange-sha256 -o Ciphers=chacha20-poly1305@openssh.com,aes256-gcm@openssh.com,aes128-gcm@openssh.com,aes256-ctr,aes192-ctr,aes128-ctr" "${PROGRAM_RETVAL_FAILURE}" + + +if [[ $num_failures == 0 ]]; then + echo -e "\n${GREENB}ALL TESTS PASS!${CLR}\n" + rm -rf -- "${TEST_RESULT_DIR}" +else + echo -e "\n${REDB}${num_failures} TESTS FAILED!${CLR}\n" +fi + +exit 0 |