The fix I did is two folds:
filters extensions (thanks to Zhujie's suggestion of a list of extensions to ban)
deletes bad files at the right place (thanks to thiscodecc's report)
Thanks: thiscodecc for the security report
Backport e4919d16ca from trunk
Removes the localhost (and 127.0.0.1) OOTB.
Allows to use it through a "multi-property" (list) in security.properties.
Conflict handled by hand in security.properties
Backport 2aa68dc4de from trunk
When you have a user with long password (greater than 256 characters) present in OFBiz and you try to log with, OFBiz return a long error message with sensitive information due to exceeding value size to store on the field UserLoginHistory.passwordUsed.
To solve this we don't return any information on the genericValue that failed and analyze the field passwordUsed to escape the case where the password set to login is create than the database field capacity.
Thanks to Daniel Elkabes <daniel.elkabes@whitesourcesoftware.com> and Hagai Wechsler <hagai.wechsler@whitesourcesoftware.com> from white source software to raise the problem.
(after discussing this on security@ofbiz.apache.org, it was decided to open a
Jira issue for that)
I've noticed that OFBiz Framework sometimes uses static initialization vectors
(IV) while creating a cipher:
https://s.apache.org/vhlbjhttps://s.apache.org/nyndk
IVs should be unique and ideally unpredictable to avoid producing the same
ciphertexts for the same plaintexts.
The issues can be fixed with something like the following:
byte[] rawIV = new byte[8];
SecureRandom random = new SecureRandom();
random.nextBytes(rawIV).
IvParameterSpec iv = new IvParameterSpec(rawIV);
jleroux: this prepends the IV in order to decrypt the plaintexts. It's similar
to what has been done for the DesCrypt class. But contrary to the DesCrypt class
I did not test.
Thanks: Artem Smotrakov
Conflicts handled by hand (a lot)
ValueLinkApi.java
(after discussing this on security@ofbiz.apache.org, it was decided to open a
Jira issue for that)
I've noticed that OFBiz Framework sometimes uses static initialization vectors
(IV) while creating a cipher:
https://s.apache.org/vhlbjhttps://s.apache.org/nyndk
IVs should be unique and ideally unpredictable to avoid producing the same
ciphertexts for the same plaintexts.
The issues can be fixed with something like the following:
byte[] rawIV = new byte[8];
SecureRandom random = new SecureRandom();
random.nextBytes(rawIV).
IvParameterSpec iv = new IvParameterSpec(rawIV);
jleroux: we prepend the IV in order to decrypt the plaintexts. I'll do
something similar in ValueLinkApi class ASAP
Thanks: Artem Smotrakov
Conflicts handled by hand:
framework/base/src/main/java/org/apache/ofbiz/base/crypto/DesCrypt.java
When trying to add a note to WorkEffort entity, I found that URL encoded
characters are unescaped, which is not expected.
e.g.:
1. Go to page: https://demo-trunk.ofbiz.apache.org/workeffort/control/EditWorkEffortNotes?workEffortId=TASK01
2. Add a note with content: https://example.com/a%20link
3. After submission, it will turned to be: https://example.com/a link
Thanks: Xin Wang for report and exchanges until solution I provided
Conflicts:
UtilHttp.java:
removes unused OOTB checkURLforSpiders for simplification
removes number of cases in setResponseBrowserDefaultSecurityHeaders
not present in R17
When creating a plugin via gradle :
./gradlew createPlugin -PpluginId=plugin-test -PpluginResourceName=PluginTest
-PwebappName=plugin-test -PbasePermission=PLUGIN_TEST
and loading all the required data and permissions, the plugin doesn't appear
in the menu.
It comes from the ComponentConfig.java file, L1229, when the permissions gets
trimmed and modified. I tried removing it locally and it fixed the problem.
Thanks: Gaetan Chaboussie
In SafeObjectInputStream.properties
Renames listOfSafeObjectsForInputStream to allowList and fixes it
Introduces a denyList
Adapts SafeObjectInputStream class to new denyList
Conflicts handled by hand
framework/base/config/SafeObjectInputStream.properties
framework/base/src/main/java/org/apache/ofbiz/base/util/SafeObjectInputStream.java
Prevents generics markup in string type names.
Conflict handled by hand
framework/base/src/main/java/org/apache/ofbiz/base/util/SafeObjectInputStream.java
This removes the inoffensive but annoying message in log (which does not show in
trunk, see https://github.com/apache/ofbiz-framework/commit/27c9180/
Not the same as in trunk, simply handling with exception Rather than returning
null, also shorten java.rmi.server to java.rmi to block more cases
The SOAP and HTTP engines are open doors to security issues.
At https://markmail.org/message/pgtjyh23bazq4s2w I proposed to comment them out
as we did for RMI in the past.
Of cause it must be clearly documented how to use them if needed.
Here is the email content:
After the recent fix for the CVE-2021-26295[1] we discussed with the security
team about the opportunity need to comment out the SOAP and HTTP engines
like we did in the past for RMI[2], this obviously for security reason.
[1] OFBIZ-12167 "Adds a blacklist (to be
renamed soon to denylist) in Java serialisation (CVE-2021-26295)"
[2] OFBIZ-6942 "Comment out RMI related
code because of the Java deserialization issue [CVE-2016-2170] "
I just put a small comment in webtools controller, it should be enough.
The tests pass
The addImageForProduct process is not directly uploading a file but renames it
before.So there is a hyphen, and possibly a pair or parenthesis in case of
duplicate files, added to the file name.
In this commit I have also added the possibility to use underscores and spaces
in the original filename
That's for (low) security reason. There are no bugs in R17 and R18 but after
reading about FREEMARKER-124 at
https://freemarker.apache.org/docs/versions_2_3_30.html
I believe we should update update Freemaker to 2.3.31 in R17 and R18
1. Navigate to URL: accounting/control/FindFinAccount
2. Click on Find button. List of fin account will be displayed.
3. Click on delete button near available balance column.
Thanks: Lalit Dashora for report, Sourabh Jain for the initial fix, Jacques for the test.
It works if we downgrade Freemarker to 2.3.28 as in R18 and even using 2.3.29,
not 2.3.30.
Handles things at the Groovy level, ie put in context, rather than creating
in Freemarker template.
It's backported, even if it's not a pb but in trunk, because it's a (low)
security issue.
In the reported case, ModelFormField::getDescription double HTML encodes the
description when both the entity and the description contain an ampersand.
A solution is to test if the description is already HTML encoded before encoding
it again. If HTML encoded then only String encodes it.
BTW I'm not sure it's useful but it's harmless, the same solution can be applied
to OFBIZ-12026 and similarly in renderableFtlFormElementsBuilder::encode. I'll
do as improvements...
Thanks: Andrew Waters for report and help in analysis
In https://markmail.org/message/pf3d3qyjfucicbj7 I initially answered:
In demo data 13 parties use <<infoString="ofbiztest@example.com">> as in
<ContactMech contactMechId="FRA01" contactMechTypeId="EMAIL_ADDRESS" infoString="ofbiztest@example.com"/>
This one is loaded last. Rather
<ContactMech contactMechId="Company" contactMechTypeId="EMAIL_ADDRESS" infoString="ofbiztest@example.com"/>
which is the 1st one, should be loaded last.
A better solution is to use "ofbizsupport@example.com" in EmailServices.xml
Indeed, fortunately the Company party also uses this infoString and using it in
EmailServices.xml does not affect the integration test and should not affect
other parts of the (only) demo data
Thanks: Emad Radwan for opiniated report :) and Pritam Kute for issue analysis
We missed to unescape EcmaScript encoded strings in
UtilCoded::checkStringForHtmlSafe, ie in all form fields using allow-html="safe"
Thanks: 牛治 <niu.zhi@zte.com.cn> for report
Conflicts handled by hand => no functional changes in code (ude to IDE setting)
framework/base/src/main/java/org/apache/ofbiz/base/util/UtilCodec.java
Follows OWASP advice on file names:
All the control characters and Unicode ones should be removed from the filenames
and their extensions without any exception. Also, the special characters such as
“;”, “:”, “>”, “<”, “/” ,”\”, additional “.”, “*”, “%”, “$”, and so on should be
discarded as well.
If it is applicable and there is no need to have Unicode characters, it is
highly recommended to only accept Alpha-Numeric characters and only 1 dot as an
input for the file name and the extension; in which the file name and also the
extension should not be empty at all
(regular expression: [a-zA-Z0-9]{1,200}.[a-zA-Z0-9]{1,10}).
So if someone needs other chars in uploaded filenames a change will be needed
Prevents too long filenames as recommended by OWASP.
Based on https://security.stackexchange.com/questions/46484/denial-of-service-when-uploading-a-file#answer-46495
I decided to not limit sizes of files. Anyway we know it's only post-auth...
When we use any title with special characters for hyperlink tag, it shows broken
string on UI, eg:
<hyperlink description="${uiLabelMap.CommonImportExport}" target="ImportExport">
should render a button with the title "Import/Export. But it shows as
Import/Export
jleroux: I have noticed that the same happens with special French characters.
So it's a general issue.
The title (description in code) is HTML encoded and that does not work for
special characters. I simply used a string encoder.
Thanks: Pradeep Choudhary for report, Sourabh Jain for a possible but not
general solution at OFBIZ-12023, ie using target-type="plain"
module should be MODULE in SecuredUpload class
Improves: add adio adn video to supported file formats in security.properties
# Conflicts handled by hand
# framework/security/config/security.properties
Handles audio and video formats supported by Tika.
Adds few new audio and video formats in seed data.
AFAIK there are no ways to embed a webshell in an audio or video file. So I did
not sophisticate the validation, just rely on Tika.
I have also fixed bugs in SecuredUpload: in isValidSvgFile and
isValidImageIncludingSvgFile
2020/08/10 the OFBiz security team received a security report by Harshit Shukla
<harshit.shukz@gmail.com>, roughly it was (quoting part of it to simplify):
<<I have identified a Remote Code Execution (RCE) Vulnerability. The reason
behind this RCE is lack of file extension check at
catalog/control/UploadCategoryImage?productCategoryId=CATALOG1_BEST_SELL&pload_file_type=category>>
Using this post-auth RCE in OFBiz demos, Harshit was able to get some AWS
credentials by uploading a webshell (based on [0]).
By security, it was then decided by the Infra and OFBiz security teams to shut
down the demos.
After discussing the elements reported with Mark J Cox (VP of ASF security team)
we in common decided that no CVE was necessary.
# Conflicts handled by hand in R18:
# applications/content/src/main/java/org/apache/ofbiz/content/data/DataResourceWorker.java
# applications/content/src/main/java/org/apache/ofbiz/content/data/DataServices.java
# applications/product/groovyScripts/catalog/category/EditCategory.groovy
# applications/product/groovyScripts/catalog/config/EditProductConfigItemContent.groovy
# applications/product/groovyScripts/catalog/imagemanagement/ImageUpload.groovy
# applications/product/groovyScripts/catalog/imagemanagement/SetDefaultImage.groovy
# applications/product/groovyScripts/catalog/product/EditProductContent.groovy
# applications/product/src/main/java/org/apache/ofbiz/product/image/ScaleImage.java
# applications/product/src/main/java/org/apache/ofbiz/product/imagemanagement/FrameImage.java
# applications/product/src/main/java/org/apache/ofbiz/product/imagemanagement/ImageManagementServices.java
# applications/product/src/main/java/org/apache/ofbiz/product/product/ProductServices.java
# build.gradle
# framework/base/src/main/java/org/apache/ofbiz/base/util/FileUtil.java
# framework/base/src/main/java/org/apache/ofbiz/base/util/HttpRequestFileUpload.java
# framework/base/src/main/java/org/apache/ofbiz/base/util/template/FreeMarkerWorker.java
# framework/security/config/security.properties
# Conflicts handled by hand in R17:
# applications/content/src/main/java/org/apache/ofbiz/content/data/DataResourceWorker.java
# applications/content/src/main/java/org/apache/ofbiz/content/data/DataServices.java
# applications/datamodel/data/seed/ContentSeedData.xml
# build.gradle
# framework/base/src/main/java/org/apache/ofbiz/base/util/HttpRequestFileUpload.java
Shuibo Ye <shuiboye@gmail.com> reported a possible arbitary file write using
webtools/control/EntitySQLProcessor.
In the "SQL Command" part, I create a table and insert some strings and
export the table to a file one sentence at a time.
PoC: CREATE TABLE "test" (string VARCHAR(80))
INSERT INTO "test" (string) VALUES ('<%= system.getProperty("user.dir") %>')
call SYSCS_UTIL.SYSCS_EXPORT_TABLE(null,'test','.\framework\webtools\webapp\webtools\default.jsp',null,'*',null)
After executing the three sentences,I successfully write the file and its url
is https://localhost:8443/webtools/default.jsp.
I fixed it preventing execution on SYSCS_UTIL.SYSCS_EXPORT_TABLE and JSP, more
could be added if necessary
This was reported to the security team by Shuibo Ye <shuiboye@gmail.com>.
We did not create a CVE because it's a post-auth "vulnerability"
Thanks: Shuibo Ye
This error appears after trying to upload a file to party content:
2020-11-06 07:48:22,488 |jsse-nio-8443-exec-7 |RequestHandler |I|
Sending redirect to:
[https://localhost:8443/partymgr/control/viewprofile?partyId=admin].
2020-11-06 07:48:29,288 |jsse-nio-8443-exec-7 |UtilObject |E|
null
java.io.NotSerializableException: org.apache.commons.fileupload.disk.DiskFileItem
# Conflicts handled by hand
# framework/base/src/main/java/org/apache/ofbiz/base/util/UtilMisc.java
# framework/webapp/src/main/java/org/apache/ofbiz/webapp/control/RequestHandler.java
For this issue (OFBIZ-8302) I reverted the point 1 of
http://svn.apache.org/viewvc?view=revision&revision=1759555
As reported by Alvaro Munoz from GH security team it's not sufficient:
<<the second part of the fix was not effective, since the attacker can close the
raw string context with a double quote and write a new attribute or even close
the macro tag and write arbitrary FreeMarker code.>>
So this removes the 2nd part and add better solution to fix the OFBIZ-8302 issue
The solution is to encode only the QueryString and to handle it correctly in
UtilHttp::getParameterMap. I must say it was not a sinecure!
# Conflicts:
# framework/base/src/main/java/org/apache/ofbiz/base/util/UtilHttp.java
# framework/widget/src/main/java/org/apache/ofbiz/widget/renderer/macro/MacroFormRenderer.java
Thanks to Alvaro's explanations, the problem was in MacroFormRenderer where, for
lookups, we retrieve _LAST_VIEW_NAME_ as a parameter without encoding it.
Got hard conflicts, all done by hand.
Thanks: Alvaro for advice
Adds an explanation about what to do when using own domain.
Some users don't look into the log so it's better to set also the exception
message in order for them to have it right at 1st glance.