Any Ideas for Addressing Security Concerns like system.getenv() ?

[phodal]

Hi, I used Kotlin Jupyter in my open source project, integration with Jupyter API. When the user run :

System.getenv()

user can get like API key and token. For example, if run this in Datalore will be:

{PATH=/opt/datalore/bin:/opt/python/envs/default/bin:/opt/datalore:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin, LOGSTASH_HOST=10.0.0.248, DISK_PORT=30081, APP_NAME=computation-agent, INSTANCE_ID=i-07e67084ee9f6dd15, LETS_PLOT_MAPTILES_KIND=vector_lets_plot, KERNEL_CLIENT_ENV=/opt/python/envs/kernel_client, MAX_HEAP_SIZE=384m, DUMP_DIR=/tmp/host/agent, EVALUATOR_MODE=container, PWD=/data/notebook_files, LOG_LEVEL=INFO, LANGUAGE=en_US:en, PYTHONPATH=/opt/datalore/python:/var/datalore/manager/.pip:/data/workspace_files, COMPUTATION_HOST=172.17.0.2, PLOTLY_RENDERER=plotly_mimetype, EVALUATOR_KERNEL_TYPE=jupyter, SQL_CELLS_API_PORT=30092, TMPDIR=/tmp, EVALUATOR_LANGUAGE=kotlin, DL_PACKAGE_MANAGER=pip, DEBIAN_FRONTEND=noninteractive, LC_ALL=en_US.UTF-8, LOGSTASH_PORT=30082, EVALUATOR_LOG_LEVEL=INFO, KOTLIN_KERNEL_SELF_CONTAINED_OUTPUTS=true, INSTANCE_TYPE=t2.medium, SHLVL=2, ANACONDA_SOURCE=/mnt/local/anaconda3, WORKBOOK_WORKING_DIR=/data/notebook_files, AGENT_SESSION_TOKEN_PATH=/data/session_token, VAR_DIR=/var/datalore, SQL_CELLS_API_HOST=10.0.0.248, AGENT_MANAGER_PORT=30090, DATALORE_USER=datalore, PYTHON_ENV=/opt/python/envs/default, LANG=en_US.UTF-8, HOST_NAME=ip-10-0-204-162, DATALOREHOME=/opt/datalore, =/opt/python/envs/kernel_client/bin/python3, KOTLIN_JUPYTER_JAVA_OPTS=, LETS_PLOT_HTML_ISOLATED_FRAME=true, DATA_ROOT=remote, DISK_HOST=disk.private.datalore.io, AGENT_RUN_TYPE=ENV, AGENT_JARS_DIR=/opt/datalore/agent, HOSTNAME=ip-10-0-204-162, AGENT_MANAGER_HOST=10.0.0.248, CHECK_ACTIVITY=true, JUPYTER_DATA_DIR=/opt/python/envs/default/share/jupyter, CONFIG_DIR=/etc/datalore, COMPUTATION_PORT=39169, HOME=/home/datalore}

I implemented a basic hook before making a request and after receiving a response. I'm now exploring ways to enhance this solution. Any suggestions?

[ileasile]

Hi! Yes, it's a valid concern. I think that the best solution is to use a safer version of JVM that doesn't execute requests to these vulnerable system methods. Another solution is response postprocessor that analyzes stream and display_data responses and removes sensitive data from them based on some heuristics

[ileasile]

However the last solution isn't safe enough because you still can execute

Runtime.getRuntime().exec(...)

[phodal]

However the last solution isn't safe enough because you still can execute

Runtime.getRuntime().exec(...)

Oops, thanks for the reminder! It will get same result.

import java.io.BufferedReader
import java.io.InputStreamReader

fun main() {
    try {
        val process = Runtime.getRuntime().exec("System.getenv()")
        val inputStream = process.inputStream
        val reader = BufferedReader(InputStreamReader(inputStream))

        var line: String?
        while (reader.readLine().also { line = it } != null) {
            println(line)
        }

        reader.close()
        process.waitFor()
    } catch (e: Exception) {
        e.printStackTrace()
    }
}