Nandish Jayaram created MADLIB-1312:
---------------------------------------
Summary: Fix potential SQL injection issues
Key: MADLIB-1312
URL: https://issues.apache.org/jira/browse/MADLIB-1312
Project: Apache MADlib
Issue Type: Improvement
Components: Deep Learning
Reporter: Nandish Jayaram
Fix For: v1.16
Based on code review comments on PR [https://github.com/apache/madlib/pull/355:]
In the madlib_keras fit function, we use the following prepare statement to
call fit transition function:
{code:java}
run_training_iteration = plpy.prepare("""
SELECT {0}.fit_step(
{1}::REAL[],
{2}::SMALLINT[],
gp_segment_id,
{3}::INTEGER,
ARRAY{4},
ARRAY{5},
$MAD${6}$MAD$::TEXT,
{7}::TEXT,
{8}::TEXT,
{9},
$1
) AS iteration_result
FROM {10}
""".format(schema_madlib, independent_varname, dependent_varname,
num_classes, seg_nums, total_buffers_per_seg, model_arch,
compile_params_to_pass, fit_params_to_pass,
use_gpu, source_table), ["bytea"])
{code}
{code:java}
$MAD${6}$MAD$::TEXT{code}
could lead to potential SQL injection issues.
Quoting related code review comments on PR #355 by [~dvaldano]:
{quote}This allows the user to inject arbitrary SQL commands, by including the
string $madlib$ in the compile or fit params. This isn't as bad as allowing
them to execute arbitrary python code, but it's something that could also
happen unintentionally and result in strange errors. We could reduce the risk
by choosing a random string, but see below for a better way of handling this.
Same issue as above, for model_arch. SQL injection possible If the user passes
a {{model_arch}} json string with {{$MAD$}} in it.
A better way to pass all of these parameters is in the same way we pass the
weights--label them as$1,$2,$3 in the SELECT statement string, and pass their
types as additional args to prepare. The values will then get passed to execute
as additional params. This is the recommended safe way to do sql queries from
python, and will do proper quoting and type casting for all of them so we don't
have to.
{quote}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
