Carol describes scenarios where the configuration options aren’t optimal, but a choice must be made.

By the time you read this, the election in the United States will be over. Many in the States view this election as having to choose between the lesser of two evils. While I’m not going to discuss the way I voted, I thought I might discuss some similar situations—where I’ve been presented with two IBM i configurations to choose from, and neither is optimal.

QSECURITY = 20 or All Profiles Are Assigned *ALLOBJ Special Authority at QSECURITY Level 40

By default, when a new profile is created at security level 20, it’s assigned *ALLOBJ special authority. So in both cases, all profiles have *ALLOBJ. In this scenario, I would prefer to have a system running at QSECURITY = 40 and all profiles assigned *ALLOBJ special authority on purpose. The reason is because of the operating system and data integrity afforded when the system runs at security levels 40 or 50. Even if my data is exposed to inappropriate modification or deletion because all users have all authority to every application object, at least the operating system can’t be inappropriately accessed or modified. And, at security level 40 or 50, data can be accessed only via an API or a command, so you know that all access is audited.

Granting Developers *ALLOBJ or *ALL Authority to Production Data

I know, I know. Developers are not supposed to have access to production systems. But I can count on one or perhaps two hands how often I’ve actually seen that to be the case. However, many organizations do limit what access developers have to production data. There’s a variety of ways organizations limit that access. Sometimes profiles are set to status of *DISABLED until access is required; sometimes developers have one profile that allows read-only access and another with access that has update authority that is supposed to be used only when data is to be modified; and other organizations use a tool that will allow elevation of authority. Regardless of the method used, when developers require access to production to modify production data, what authority should they have—*ALLOBJ special authority or *ALL to production data? My preference is to give them *ALL authority to production data. With *ALLOBJ, they have all authority to all objects on the system and you cannot block their access. With *ALL authority, it’s limited to application objects. There are a couple of ways to grant them *ALL authority. You can either grant them *ALL authority to an authorization list securing the application data or add them to the group that owns the application. Either method limits the developers’ elevated access and doesn’t automatically give them all authority to everything on the systems.

QCRTAUT = *ALL or *ALLOBJ Special Authority

The QCRTAUT system value defines the *PUBLIC authority settings when most objects are created. You can override this value when you create the object or in the library description where the object is being created. But if you do nothing, most objects will be created with *PUBLIC *ALL. Let’s look at this scenario. When you create or change a profile to have *ALLOBJ special authority, the effect is virtually immediate. The next time the user signs onto the system, they have all authority to every object on the system. Contrast that with changing QCRTAUT from the default of *CHANGE to *ALL. That change has no effect on existing objects. For example, if the *PUBLIC authority setting of your employee master database is *EXCLUDE, it will remain *EXCLUDE. In addition, even after an object is created as *PUBLIC(*ALL), it can be changed. When *ALLOBJ is assigned to a user profile, you cannot prevent users from accessing objects. That’s because *ALLOBJ is the first thing that’s checked when the system runs its authority-checking algorithm. So if you haven’t guessed, my preference would be to have QCRTAUT set to *ALL rather than all users configured with *ALLOBJ special authority.

Auditing Not Configured or No Audit Journal Receivers on the System

When the QAUDCTL system value is *NONE, auditing is not active. However, auditing may be active and you may be logging every action possible, but if you keep only the current journal receiver on the system, it’s not much better than having no auditing active, especially on a busy system. That said, I would still take auditing configured and active rather than not configured. In this situation, I would try to convince the organization to keep, at the bare minimum, 24 hours of information on the system. Better would be to keep three days of information so that if an event occurs on a Friday that needs investigation, you’d have those receivers on the system. But even if you can’t keep any receivers other than the current one, as long as you’re saving the receivers prior to deletion, you’ve got the information available should you need it to perform forensics.

QMAXSIGN Set to *NOMAX or QPWDEXPITV Set to *NOMAX

QMAXSIGN controls how many times an organization allows users to guess their password before some action is taken (the profile and/or the device is disabled). The QPWDEXPITV system value determines how often a user’s password must be changed. Setting QMAXSIGN to *NOMAX means that users can try indefinitely to come up with the right password. That may not be such a big deal when it’s a legitimate user, but think about the ramifications when the person attempting to sign on is a hacker or someone inside your organization who is trying to gain access with a more powerful profile (such as QSECOFR) or sign on with your CEO’s profile and do harm? When QPWDEXPITV is set to *NOMAX, users are never required to change their password. Assuming no one ever guesses someone else’s password and no malware has been installed somewhere in your network to sniff cleartext user ids and passwords, this is not as dangerous as giving users an unlimited number of attempts to come up with a correct user id-password combination.

Keep Inactive Profiles on the System or Suffer Application Issues

On more than one occasion, I’ve seen application issues arise when deleting inactive profiles. In one situation, a web application was written to look for the presence of a user profile. The application never used the profile for authentication or to run a job. But if the profile wasn’t present, the user couldn’t sign on to the web app. (Don’t get me started on how bad this design is. That discussion must wait for another article!) In another situation, when running a historical report of transactions that occurred within the application, the profile’s text description was pulled out of the profile in real time. If the profile no longer existed when the report was run, the log report wouldn’t format correctly and didn’t include the profile name. Again, the insanity of that design will have to be discussed another time. In the other scenario, leaving inactive profiles on a system is an unnecessary risk. If the profile remains enabled, someone can attempt to guess the password and will be on using a profile that is likely not audited and not expected to perform actions and may have additional capabilities beyond their own. Even if the profile’s disabled, I have seen numerous occasions when all profiles on the system were re-enabled (especially during a role-swap.) In addition, leaving inactive profiles on the system causes your Save Security Data (SAVSECDTA) process to run longer. And worse, if you have to perform a restore of your system, the Restore User Profile (RSTUSRPRF) command will have to put back all of those unused profiles. That said, if an application doesn’t work properly, what choice do you have? If you are forced to leave inactive profiles on the system, I recommend that you do more than set the profile to be status of disabled. Consider also setting the following attributes:

• PASSWORD(*NONE)
• INLPGM(*NONE)
• INLMNU(*SIGNOFF)
• ATNPGM(*NONE)
• SPCAUT(*NONE)
• GRPPRF(*NONE)
• LMTCPB(*YES)
• PWDEXPITV(*SYSVAL)
• MAXSTG(0)
• STATUS(*DISABLED)

These settings take all opportunities to make sure the profile cannot be used.

The Best of the Worst

None of these scenarios are good, but I hope that you’ve enjoyed the discussion of how I chose the “lesser of two evils” when it comes to various IBM security scenarios.