Tag Archives: aem

AEM Security: How to secure the AEM application?

Leave a reply

Overview

There is a set security practice followed by every development team in Adobe experience manager ( i.e AEM) CMS technology. And, Most of these are pretty straightforward suggested by the Adobe as best practices however there are many other security issues which have equal importance.

So, Let’s begin to know how to secure your application by putting right rules in your AEM environment.

All other recommendations from the open web application security project(i.e OWASP) should be applied. Below recommendations are very specific to AEM technology & AEM infrastructure.

There are many problems which are unknown to the AEM Solution provider & putting the whole thing at risk. I would like to state one of the examples here to showcase the security problems in AEM.

Use below Google Query to find out if your author instance is indexed by the google or not. I have used a very basic query in google. Try it, you would surprise to see how many author instances which are open to exploits. You might be wondering how to login in those authors. That is fairly easy once you know who has authored the pages.

Google Query: inurl:aemauthor

AEM Author Security:

First & foremost, Make sure your AEM author instance isn’t searchable by the search engine & It is not accessible outside of Intranet without VPN. Follow some author security guidelines below:

  • Keep robots.txt for all your domains including the authoring environment. make sure Google does not index author domain.
  • Enable HTTPS in AEM Author.
  • Changing Admin password in every AEM instance (i.e server).
  • Create groups for assigning access & follow the least privilege principle. Basically, Instead of denying on many hierarchies just allow what individual group needs.
  • Create a separate replication user to use in replication agent configuration. Admin should not be used for replicating anywhere.
  • Limit the number of users in admin groups.
  • Web dev, CRX explorer & CRXDE in prod author should be disabled or should be limited to certain users.

AEM Publish Security

Same as AEM author, publish instances should not be accessible to an outside of the intranet & connections to web servers, author etc should be internal connections. The most important thing to handle in publish security is to handle requests inputs & use proper request sessions. Serving requests with admin session or privileged user is a big problem. 

Assume some data you have to read & anonymous user does not have permission to that then avoid using admin session. Have a dedicated user for that to read/write the content for certain requests. Follow other guidelines respect with AEM Publish security:

  • Anonymous permissions should be checked & make sure not every directory accessible to the anonymous user. Even in etc design, There should be proper permission setup in cloud services etc.
  • Apache Sling Referrer Filter must be configured to handle unwanted publish requests.
  • The cross-site forgery framework should be enabled to filter requests.
  • All default tools (Crx explorer, Crxde, WebDev) etc should be disabled.
  • No one should be able to access publish server directly. Also should not be able to install packages directly.

Dispatcher security

When anyone thinks of AEM security, most of us just think of rules & filters in dispatcher.any configuration file. But, There are many more use cases where things are not pretty if you have not taken care of security:

  • Do not have dispatcher flush agent configured from AEM author. And if it is enabled then have https call for flushing cache. Otherwise, author flush agent exposes to your web server IP & credentials.
  • Limit the request headers information. Request headers are passed in every request to AEM publish based on dispatcher configuration.
  • Do not allow cross-origin requests. Set the SAME origin header at the web server level.
  • Proper input validation should be done in POST Requests & dispatcher filter should allow only certain POST requests.
  • Caching of selectors & URL extensions should be defined. Not every selector or extension should be cacheable. DOS or DDOS attacks are very easy to do in AEM application.
  • Website URL’s should not expose internal directories.

Final thought

We have to secure the infrastructure & security of important environments. Once you have security author, publish & proper dispatcher configuration, you would have a better chance to protect your application. Application security is another aspect follow the below links for Adobe recommendation.

Developer Opinion: A few basic problems in Sightly template

Leave a reply

AEM Sightly has been a big buzz in AEM technology & there are thousands of articles written on it. Most interesting things to notice that most of the articles only talk about syntax & how to use sightly code. This is great for the Developers to just re-use code.

In this post, I would like to raise some important questions which I have not found in any articles or posts. If anyone finds answer & explanation, I would thank him if he/she let me know in comments. Will keep recording other Sightly problems but for now, Here are some of them:

Why else Condition not supported?

As far as I know, most of the languages support else condition whether language is scripting, template or OOPS except AEM Sightly template language. We don’t know what super logic & architecture drawbacks it has to support else condition. One argument is that Adobe does not want else condition in HTML tags. Fair enough but what problem you may have with ‘else’. I understand that Nobody wants business logic to be in HTML. But, That is the best practices Adobe can recommend.

Let’s understand a scenario for the clarity & why else is equally important. 

Let’s say, Based on some authoring checkbox, You would like to serve different HTML Content (i.e two types of DOM structure). If you don’t have else then developer has to put if with boolean condition & another if with the negative condition.

<sly data-sly-test=${IsFlagChecked}> // show checked content </sly>

<sly data-sly-test=${! IsFlagChecked}> // show non checked content </sly>

The workaround with NOT operator is fine & by this logic, someone might think that else statement isn’t required at all when there is a workaround. My argument would be workaround should not be objective when you design a new template language. There must be some solid use case not to support that.

Why no argument allowed in get method?

Let’s consider a scenario where you want to get output from Sling Model in different instances with the HTML but output varies based on parameters passed to the method. Can you pass the parameters to the GET method mentioned in USE class? The answer is NO. 

Passing parameters from Sightly is only allowed at the time of USE class initialization. It means you should know all types of inputs to different methods before calling the GET method. And, Sightly is forcing you to have global variables initialized when USE class gets initialized. In general, Not a good idea to keep every business logic with global variables. It seems to me that AEM works good because the objective is to generate HTML & cache it. Nothing else.

Why No method overloading not supported?

When you have a common abstract USE class with a bunch of common defined methods & you wish to overload that method with a different parameter, that is not the possible at all. You can’t call your overloaded method. Your USE class must have their own method & supporting GETTER method for the global variables. Basically, end up creating extra codes which may not be required at all.

Again, I would not deny that this can’t be solved with work around. However, There must be some explanation for not following concept which we have been practising & grow up learning many common design principles. 

Final Thoughts

I believe there may have some good reasons to do the way it is now. But, logically it doesn’t seem right to me. And, my intention is to get those answer through this article. Hoping that someone might know the answers.

AEM Solution: The easiest way to copy content from one AEM to another.

7 Replies

Moving Content in AEM is a big task regularly. In my personal opinion, it is big task for everybody. Let me try to explain in details. Let’s consider a scenario where you want to move content from one AEM environment to another. The easy thing is to do to use AEM Package manager. That is good. And just build a package from one AEM, download it & install somewhere else. Easy process? You may think it is but it is not. From the Business perspective, the Package Manager tool totally sucks & for the following reasons:

Lack of basic features in Package Manager:  There are many basic features missing. Some of them are:

  • No way you can schedule the content package as a whole. And, if 100 pages to be scheduled then Each individual pages must be scheduled to replicate them.
  • No way you could upload the individual pages content from one environment to another if individual pages are the parent pages in the content hierarchy. All the content has to be overridden.
  • Not easy to revert the certain content if installed by the package manager. either whole content or nothing can be reverted.

Not easy to use by the Non-Technical Person:  Authoring team must have a working knowledge of package manager tool. I know you might think working knowledge? My answer would be YES. Someone needs to know how to upload, build, install, download & uninstall etc. And needs access to the packages when someone can misuse it.

Time-consuming & does not work in most of the cases: Downloading from one environment & uploading in another is very old fashion & time-consuming. For heavy content like size GB’s, It does not even work. 

So, Here are the list of possible Solutions:

  • TWC Grabbit is one of them. It was developed by one of our team members however not sure if it is working in all the AEM versions. It has so many dependencies & Needs to install & managed in source & destination. But it was a quite good one.
  • AEM Package Manager Out of the box.
  • Copy whole source CRX-QUICKSTART folder & override the destination: Not a feasible option if the content has to be moved to production from stage or from stage to prod. Also not a solution if you want to move the only fewer pages or images. However, Not bad solution for Dev & QA but comes with lots of maintenance once the content is overridden.

The most easiest way move content regularly

You need to have just two things: Have a servlet in source code & Configure destination replication agent. You can see below it.

Pros/Cons of this solution:

  • First, a good thing is that it is pretty easy & you can replication any JCR path. Include a content package, one page/child pages, one image/set of images. if you replicate a content package then no need to install in the destination environment. And, Helpful when you just need some pages in your QA or dev from the stage. Not whole content.
  • No dependency. No installation. Just one servlet, replication agent. And, using out of the box solutions. 
  • Pretty extensible. You can build fancy UI out of it & make it a tool out of it.
  • Cross-environment replication & replication only for content movement. Any environment can be a source or destination. Having a separate replication agent just for copying content does not cause any replication queue issue.
  • Cons is it is still using replication API & not any fancy third-party solution.

NOTE: I have tool build around which solves all the issue a content package has. But, not yet sure if I could simply provide source code here. However, let me know if you need some help or idea to understand the full solution.

Hit this URL from a browser after your servlet & agent is done:  http://localhost:4502/bin/support/content/publisher?path=etc/packages/abc.zip&destEnvName=QA&publishChildNodes=true.&nbsp; publishChildNodes is required when you want to publish child nodes also.

Agent Configuration in AEM Author

Replication authoring – Nothing different from other replication agent except Triggers configuration. Do the same as you see in the snapshot.

Replication Request Handler

import com.day.cq.replication.*;
import org.apache.commons.lang3.StringUtils;
import org.apache.felix.scr.annotations.Component;
import org.apache.felix.scr.annotations.Reference;
import org.apache.felix.scr.annotations.sling.SlingServlet;
import org.apache.http.HttpStatus;
import org.apache.sling.api.SlingHttpServletRequest;
import org.apache.sling.api.SlingHttpServletResponse;
import org.apache.sling.api.resource.Resource;
import org.apache.sling.api.servlets.SlingAllMethodsServlet;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import javax.jcr.Session;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;

/**
 * Sample URL http://localhost:4502/bin/support/content/publisher?path=etc/packages/abc.zip&destEnvName=QA&publishChildNodes=true
 */
@SlingServlet (paths = "/bin/support/content/publisher",
 methods = "GET", metatype = true, label = "Content publisher to publish content across environments")
public class PackagePublisher extends SlingAllMethodsServlet {
    private static final Logger LOGGER = LoggerFactory.getLogger(PackagePublisher.class);

    @Reference
    private Replicator replicator;
    private List<String> activatedPathsList;
    @Override
    public final void doGet(final SlingHttpServletRequest request, final SlingHttpServletResponse response) throws IOException {
  String requestPath = request.getParameter("path");
  String publishChildNodes = request.getParameter("publishChildNodes");
  final String destEnvName = request.getParameter("destEnvName");
  if (StringUtils.isNotBlank(requestPath) && StringUtils.isNotBlank(destEnvName)) {
      activatedPathsList = new ArrayList<String>();
     Session userSession = request.getResourceResolver().adaptTo(Session.class);
  ReplicationOptions replicationOptions = new ReplicationOptions();
 AgentFilter agentFilter = new AgentFilter() {
    public boolean isIncluded(Agent agent) {
 if(agent.getId().toLowerCase().contains(destEnvName.toLowerCase())) {                   return true;
                    }
                    return false;
                }
            };
            replicationOptions.setFilter(agentFilter);
            LOGGER.info("replication starting ");
            try {
                replicator.replicate(userSession, ReplicationActionType.ACTIVATE, requestPath, replicationOptions);
                Resource childResource = request.getResourceResolver().getResource(requestPath);
                if ("true".equalsIgnoreCase(publishChildNodes)) {
                       publishChildPages(childResource, userSession, replicationOptions);
                }
                for (String path: activatedPathsList){
                    LOGGER.info("Activate paths" + path );
                }
                response.setStatus(HttpStatus.SC_OK);
                response.getWriter().print("given path is replicated to given environment. Check in destination env.");
            } catch (ReplicationException e) {
                response.setStatus(HttpStatus.SC_BAD_REQUEST);
                response.getWriter().print("Check Parameters. Also check author replication agents for " + destEnvName);
                e.printStackTrace();
            }catch (Exception ex){
                response.setStatus(HttpStatus.SC_BAD_REQUEST);
                response.getWriter().print("Something was wrong!!");
            }
        } else{
            response.setStatus(HttpStatus.SC_BAD_REQUEST);
            response.getWriter().print("Parameters are not passed.");
        }
    }

    private void publishChildPages(Resource childResource, Session userSession,
                                   ReplicationOptions replicationOptions) throws ReplicationException {
             if (childResource != null) {
                Iterator<Resource> itr = childResource.listChildren();
                while (itr.hasNext()) {
                    Resource temp = itr.next();
                    if (!temp.getPath().contains("rep:policy") && !temp.getPath().contains("jcr:content")) {
                        if (temp.hasChildren()) {
                            publishChildPages(temp, userSession, replicationOptions);
                        }
                        activatedPathsList.add(temp.getPath());
                        replicator.replicate(userSession, ReplicationActionType.ACTIVATE, temp.getPath(), replicationOptions);
                    }
                 }
            }
    }
}

Final Thought

I found it very easy in day to day work when you want to move content here & their. However, if there is any confusion & question. leave a comment. will respond asap. thanks.

AEM Solution: How to create an AEM Custom global object?

1 Reply

Overview

As an AEM developer, you know that there are a set of global objects available in AEM at the view layer. Whether you can use JSP or New template framework (HTL). Most commons global objects are currentPage, current resource, current design object etc.

These set of objects are initialized by the AEM application by default & available to use them in every component.

Do you really need custom object?

With given architecture of AEM & Sling framework to resolve the resource, there is a situation when you need your own custom object to be available across the application components. Just like global objects.

Let’s consider a multi-brand site where you need some brand-specific information, Objects or JSON etc should be available at each brand application level. It means object should be available at every component.

General Solution

General practice is that have a common bean/Use class gets initialized at each component level. Performance wise, initializing a bean at every component level, isn’t that bad. Because most of the content gets cached. However, It is an issue when you have got that information/object through service. You can’t make a web service from every component. You could still do that if you are an architect & developer has an alternative solution. So let’s see an alternative solution.

Alternative Solution

An alternative solution is pretty simple. Just like any other AEM global objects, keep your custom object available at every component by default. No initialization of any objects. Make sense? let’s see the following steps to create global custom objects.

  • Create a new class with an appropriate name (hard to find the appropriate name but try it). 
  • Make sure new class is registered as Component in OSGI Console.
  • Sling Framework provides an interface BindingValuesProvider. Implement this interface. This interface provides addBindings(Bindings bindings) method.
  • Bindings parameter works like a map. You need to put key & value in it. For instance, bindings.put(“myObject”, “valueOfCustomObject”);
  • Now, “myObject” can be used as a global variable anywhere in the components.

Sample src Code

# CustomGlobalObject
import org.apache.felix.scr.annotations.Component;
import org.apache.felix.scr.annotations.Reference;
import org.apache.felix.scr.annotations.Service;
import org.apache.sling.scripting.api.BindingsValuesProvider;
import javax.script.Bindings;
@Component
@Service
public class CustomGlobalObject implements BindingsValuesProvider {
    @Reference
    IMyService iMyService;
    @Override
    public void addBindings(Bindings bindings) {
        bindings.put("mylist", iMyService.getGlobalObject());
    }
}
#Service Interface
import java.util.List;
public interface IMyService {
    List getGlobalObject();
}
#Service Interface implementation

import org.apache.felix.scr.annotations.Component;
import org.apache.felix.scr.annotations.Service;
import java.util.ArrayList;
import java.util.List;
@Component
@Service
public class MyService implements IMyService{
    @Override
    public List<String> getGlobalObject() {
        List list = new ArrayList();
        list.add("a"); list.add("b"); list.add("c");
        return list;
    }
}
# Entry in Sightly HTML component
Custom object: ${mylist}
#Output
Custom object: a,b,c

Final Thought: 

There could be other good solutions to solve the problem which I described in this post. Would love to hear others ideas. Don’t hesitate to share your thoughts in the comment section. For any doubt or question, leave a comment. will do our best to answer your questions. Thanks in advance.

References

AEM Solution: AEM OSGi Config Resolution Order

1 Reply

Overview

In this post, will talk about simple concept what is OSGi (i.e open service gateway interface) resolution order & what does this order is different when AEM starts & when something we change from OSGi console.

OSGI Config Resolution Order: at Startup vs Runtime

AEM Apache Felix based OSGi runtime environment(aka system console) loads the configurations in two different ways and makes them available to the application. Read the section “How these configs are resolved?” in this post How AEM OSGi works?

The same OSGi Apache Felix framework loads configuration runtime different. The following order of precedence applies: 

  • If you have modified any config directly from system console (i.e AEM OSGi admin console) then AEM creates another .config file and this file gets precedence over your XML file, apps or libs.
  • If you have made changes in config file under apps and the same config has not been modified from system console then apps modified config would take the precedence over libs and will available to the application immediately.
  • If you have modified any config under libs & it is not overridden at the apps level & it is not modified from system console then modified config will take the precedence.

Config Resolution Order with multiple run modes:

For run mode specific configurations, multiple run modes can be combined. For example, you can create configuration folders in the following style:

/apps/*/config.<runmode1>.<runmode2>/

Configurations in such folders will be applied if all run modes match a run mode defined at startup. For example, if an instance was started with the run modes author,dev,emea, configuration nodes in /apps/*/config.emea, /apps/*/config.author.dev/ and /apps/*/config.author.emea.dev/ will be applied, while configuration nodes in /apps/*/config.author.asean/ and /config/author.dev.emea.noldap/ will not be applied.

If multiple configurations for the same PID are applicable, the configuration with the highest number of matching run modes is applied.

For example, if an instance was started with the run modes author,dev,emea, and both /apps/*/config.author/ and /apps/*/config.emea.author/ define a configuration for
com.day.cq.wcm.core.impl.VersionManagerImpl, the configuration in/apps/*/config.emea.author/ will be applied.

This rule’s granularity is at a PID level. You cannot define some properties for the same PID in/apps/*/config.author/ and more specific ones in /apps/*/config.emea.author/ for the same PID.
The configuration with the highest number of matching run modes will be effective for the entire PID.