Garak integration (#306)

* added code for scanning with Garak
* created a garak config template
* added a separate Containerfile to build an image that includes Garak
* updated README to introduce this feature

README.md

@@ -6,8 +6,9 @@ RapiDAST (Rapid DAST) is an open-source security testing tool that automates DAS
 
 RapiDAST provides:
 
-- Automated HTTP/API security scanning using ZAP
-- Kubernetes operator scanning using OOBTKUBE
+- Automated HTTP/API security scanning leveraging ZAP
+- Automated LLM AI scanning leveraging Garak
+- Kubernetes operator scanning leveraging OOBTKUBE
 - Automated vulnerability scanning using Nessus (requires a Nessus instance)
 - Command-line execution with yaml configuration, suitable for integration in [CI/CD pipelines](./examples/)
 - Ability to run automated DAST scanning with pre-built or custom container images
@@ -131,6 +132,7 @@ See templates in the [config](./config/) directory for rapidast configuration ex
 - `config-template-zap-long.yaml` : describes a more extensive use of ZAP (all configuration options are presented)
 - `config-template-multi-scan.yaml` : describes how to combine multiple scanners in a single configuration
 - `config-template-generic-scan.yaml` : describes the use of the generic scanner
+- `config-template-garak.yaml` : describes the use of the Garak LLM AI scanner
 
 See [here](./examples/) for examples on how to run RapiDAST in various CI/CD pipelines.
 
@@ -470,9 +472,21 @@ scanners:
       - 127.0.0.1
 ```
 
+#### Garak
+
+Garak is an LLM AI scanner developed by NVIDIA. See https://github.com/NVIDIA/garak for more information.
+
+The following is an example to launch a scan:
+```yaml
+scanners:
+  garak:
+    model_type: huggingface
+    model_name: gpt2
+```
+
 #### Generic scanner
 
-RapiDAST can run other scanning tools as well as ZAP. It is possible to request RapiDAST to run a command and process stdout results, using the `generic` plugin.
+In addition to the scanners mentioned above, RapiDAST can run any other scanning tools. It is possible to request RapiDAST to run a command and process stdout results, using the `generic` plugin. One use case is to run your own tools or scripts and export the results to Google Cloud Storage.
 
 The following is an example to run a command or a tool in the host where a RapiDAST scan runs:
 

config/config-template-garak.yaml

@@ -0,0 +1,29 @@
+config:
+  # WARNING: `configVersion` indicates the schema version of the config file.
+  # This value tells RapiDAST what schema should be used to read this configuration.
+  # Therefore you should only change it if you update the configuration to a newer schema
+  configVersion: 6
+
+  # all the results of all scanners will be stored under that location
+  # base_results_dir: "./results"
+
+# `application` contains data related to the application, not to the scans.
+application:
+  shortName: "garak-test-1.0"
+
+# `scanners' is a section that configures scanning options
+scanners:
+  garak:
+    model_type: huggingface                         # required, e.g. hugginngface, openai, rest
+    model_name: gpt2                                # optional, but a specific model type requires a model name or path
+    #generators:                                    # optional, providing more options for the selected model type, e.g. RestGenerator
+    #  rest:
+    #    RestGenerator:
+    #      uri:
+    #      method:
+    #      headers:
+    #      response_json_field:
+    #      req_template_json_object:
+    #      request_timeout: 60
+    #probe_spec: all                               # default: all, or a list of probes like "probe1,probe2"
+    #garak_executable_path: /usr/local/bin/garak    # default: /usr/local/bin/garak

config/schemas/6/rapidast_schema.json

@@ -634,6 +634,31 @@
                             "scan",
                             "server"
                         ]
+                    },
+                    {
+                        "type": "object",
+                        "description": "Garak Scanner",
+                        "properties": {
+                            "model_name": {
+                                "type": "string"
+                            },
+                            "model_type": {
+                                "type": "string"
+                            },
+                            "generators": {
+                                "type": "object",
+                                "properties": {}
+                            },
+                            "probe_spec": {
+                                "type": "string"
+                            },
+                            "garak_executable_path": {
+                                "type": "string"
+                            }
+                        },
+                        "required": [
+                            "model_type"
+                        ]
                     }
                 ]
             }

containerize/Containerfile.garak

@@ -0,0 +1,113 @@
+#####
+# Build RapiDAST image with Garak LLM AI scanner: https://github.com/NVIDIA/garak
+#####
+
+# Prepare dependencies
+FROM registry.access.redhat.com/ubi9-minimal AS deps
+
+# Indicate if dependencies were prefetched using Cachi2
+# They must be located at /cachi2/output/deps
+ARG PREFETCH=false
+
+# These versions should be consistent with the listed in the artifacts.lock.yaml file
+ARG ZAP_VERSION=2.15.0
+ARG FF_VERSION=128.6.0esr
+ARG K8S_VERSION=1.32.1
+ARG TRIVY_VERSION=0.59.0
+
+ARG DEPS_DIR=/tmp/deps
+ARG ZAP_FILE=$DEPS_DIR/ZAP.tar.gz
+ARG FF_FILE=$DEPS_DIR/firefox.tar.bz2
+ARG TRIVY_FILE=$DEPS_DIR/trivy.tar.gz
+ARG KCTL_FILE=$DEPS_DIR/kubectl
+
+RUN microdnf install -y tar gzip bzip2 java-11-openjdk nodejs
+
+RUN mkdir "${DEPS_DIR}" /tmp/node_modules && if [ "$PREFETCH" == "true" ]; then \
+    echo "PREFETCH is true: Copying dependencies from /cachi2/output/deps..." && \
+    cp -r /cachi2/output/deps/generic/* "$DEPS_DIR"; \
+  else \
+    echo "PREFETCH is false: Downloading dependencies from remote sources..." && \
+    curl -sfL "https://github.com/zaproxy/zaproxy/releases/download/v${ZAP_VERSION}/ZAP_${ZAP_VERSION}_Linux.tar.gz" -o "$ZAP_FILE"; \
+    curl -sfL "https://releases.mozilla.org/pub/firefox/releases/${FF_VERSION}/linux-x86_64/en-US/firefox-${FF_VERSION}.tar.bz2" -o "$FF_FILE"; \
+    curl -sfL "https://dl.k8s.io/release/v${K8S_VERSION}/bin/linux/amd64/kubectl" -o "$KCTL_FILE"; \
+    curl -sfL "https://github.com/aquasecurity/trivy/releases/download/v${TRIVY_VERSION}/trivy_${TRIVY_VERSION}_Linux-64bit.tar.gz" -o "$TRIVY_FILE"; \
+  fi
+## ZAP, build and install scanners in advance (more scanners will be added)
+RUN mkdir /opt/zap && \
+  tar zxvf "$ZAP_FILE" --strip-components=1 -C /opt/zap && \
+  ### Update add-ons
+  /opt/zap/zap.sh -cmd -silent -addonupdate && \
+  ### Copy them to installation directory
+  cp /root/.ZAP/plugin/*.zap /opt/zap/plugin/
+
+## Firefox, for Ajax
+RUN mkdir -p /opt/firefox && \
+  tar xjvf "$FF_FILE" -C /opt/firefox
+
+## kubectl
+RUN install -o root -g root -m 0755 "$KCTL_FILE" /usr/local/bin/kubectl
+
+## Trivy (https://github.com/aquasecurity/trivy/)
+RUN mkdir /tmp/trivy && \
+  tar xzvf "$TRIVY_FILE" -C /tmp/trivy && \
+  chmod +x /tmp/trivy/trivy
+
+## redocly (https://github.com/Redocly/redocly-cli)
+COPY package.json package-lock.json /tmp/redocly/
+RUN mkdir -p /tmp/redocly/node_modules && if [ "$PREFETCH" == "true" ]; then \
+    npm install --offline --prefix /tmp/redocly; \
+  else \
+    npm install --prefix /tmp/redocly; \
+  fi
+
+# Copy artifacts from deps to build RapiDAST
+FROM registry.access.redhat.com/ubi9-minimal
+
+COPY --from=deps /opt/zap /opt/zap
+COPY --from=deps /opt/firefox /opt/firefox
+COPY --from=deps /usr/local/bin/kubectl /usr/local/bin/kubectl
+COPY --from=deps /tmp/trivy/trivy /usr/local/bin/trivy
+COPY --from=deps /tmp/redocly/node_modules /opt/redocly/node_modules
+
+ENV PATH $PATH:/opt/zap/:/opt/rapidast/:/opt/firefox/
+
+## RapiDAST
+RUN mkdir /opt/rapidast
+COPY ./rapidast.py ./requirements.txt /opt/rapidast/
+COPY ./scanners/ /opt/rapidast/scanners/
+COPY ./tools/ /opt/rapidast/tools/
+COPY ./exports/ /opt/rapidast/exports/
+COPY ./configmodel/ /opt/rapidast/configmodel/
+COPY ./utils/ /opt/rapidast/utils/
+COPY ./config/ /opt/rapidast/config/
+
+### Add generic tools in the PATH
+COPY ./scanners/generic/tools/convert_trivy_k8s_to_sarif.py /usr/local/bin/
+
+### Overload default config (set 'none' as default container type)
+COPY ./containerize/container_default_config.yaml /opt/rapidast/rapidast-defaults.yaml
+
+### Add /opt/{zap,rapidast}/ to the PATH (for any user and future user)
+COPY ./containerize/path_rapidast.sh /etc/profile.d/rapidast.sh
+
+### Install RapiDAST requirements, globally, so that it's available to any user
+RUN microdnf install -y --setopt=install_weak_deps=0 java-11-openjdk shadow-utils dbus-glib procps git nodejs npm && \
+  microdnf install -y gtk3 python3.12 rust cargo && \
+  python3.12 -m ensurepip --upgrade && \
+  pip3.12 install --upgrade pip && \
+  pip3.12 install --no-cache-dir -r /opt/rapidast/requirements.txt && \
+  pip3.12 install -U garak --no-cache-dir && \
+  microdnf clean all -y && rm -rf /var/cache/dnf /tmp/*  && \
+  ln -s /opt/redocly/node_modules/@redocly/cli/bin/cli.js /usr/local/bin/redocly
+
+RUN useradd -u 1000 -d /opt/rapidast -m -s /bin/bash rapidast && \
+  chown -R 1000 /opt/rapidast && \
+  # OpenShift runs containers with arbitrary user ids, belonging to root group
+  chgrp -R 0 /opt/rapidast && \
+  chmod -R g=u /opt/rapidast
+
+USER rapidast
+WORKDIR /opt/rapidast
+ENV HOME /opt/rapidast
+ENTRYPOINT ["./rapidast.py"]

scanners/garak/garak-config-template.yaml

@@ -0,0 +1,39 @@
+---
+system:
+    verbose: 0
+    narrow_output: false
+    parallel_requests: false
+    parallel_attempts: false
+    lite: true
+    show_z: false
+
+run:
+    seed:
+    deprefix: true
+    eval_threshold: 0.5
+    generations: 5
+    probe_tags:
+
+plugins:
+    model_type:
+    model_name:
+    probe_spec: all
+    detector_spec: auto
+    extended_detectors: false
+    buff_spec:
+    buffs_include_original_prompt: false
+    buff_max:
+    detectors: {}
+    generators: {}
+    buffs: {}
+    harnesses: {}
+    probes:
+        encoding:
+            payloads:
+                - default
+
+reporting:
+    report_prefix:
+    taxonomy:
+    report_dir: garak_runs
+    show_100_pass_modules: true