Shodan and SERC Audits

On SERC’s Open Forum Webinar on Monday, January 25, they had a presentation on the Shodan search engine. I wasn’t able to participate in the webinar, but a co-worker did, and the webinar seemed to be based on this one-page document SERC released on January 4, “Shodan: What SERC Registered Entities Need to Know.” SERC said on the webinar that the SERC audit teams will be using Shodan as part of their audits.

Shodan (www.shodan.io) calls itself the “world’s first search engine for Internet-connected devices.” If you haven’t had a chance to explore Shodan yet, you really should go spend a half-hour or so exploring it and seeing what is possible with it. It’s been in the news over the last couple days because they added a new section that allows paid users to browse Internet-connected webcams. You don’t have to be a paid user to use the site, although the paid accounts do have some added capabilities. Shodan also has a section dedicated to Industrial Control Systems, where a user can get tips on how to find ICS. One of their examples is that if you search for ‘title:”xzeres wind”‘, then you can find wind turbines.

SERC’s one-page is a pretty good overview of what Shodan does. One key point that they make is that when somebody uses Shodan, they aren’t actually contacting the target. Shodan has already gotten the information about what devices are on the Internet, and all the user of Shodan is doing is searching this database. SERC calls out two things which an attacker could use Shodan for. First, if an attacker has a vulnerability they are prepared to use, they could use Shodan to find devices which have that vulnerability. As an example, the first alert on ICS-CERT’s website when I checked it today was ICS-ALERT-15-225-01A for Rockwell Automation 1769-L18ER PLC. This alert talks about there being public proof-of-concept exploit code. If you search Shodan for “1769-L18ER/A”, you find 65 results. An attacker could use that list to find possible targets for an attack.

If you’re unlucky (probably the wrong word, because you’re doing something wrong, not just being unlucky) to have one of the devices somebody is targeting on the list that gets returned to the user, you might get targeted. There isn’t much that knowledge of this type of attack using Shodan can actually do for you, though. It’s probably unrealistic for a company to try to go search for all the devices they have in their plant, and even if they did, almost everythiing they found on Shodan would be a false positive since any one result isn’t likely to be from your specific organization.

The other way that SERC points out that an attacker could search an organization’s domain name. This would be used in a more targeted approach. Using the domain of a large energy company, there were 40 results returned by Shodan.

This led mostly to some 302 Redirects and mail servers whose banner message said that the Shodan crawler was not allowed to access them. There weren’t any any ICS protocols in any of the results, as you can see in the “Top Services” section, so this company seems to be in pretty good shape, at least in terms of what Shodan could see in a simple search.

SERC said that they recommend that,

all its registered entities perform a Shodan search for their own domain name(s). For each device found, consider the following questions:
Is it truly necessary for this device to be external-facing?
Have all applicable security patches been installed on this device?
Have unneeded accounts and old passwords been removed/changed?
What security measures are in place on the device (e.g., anti-virus, host-based IPS, firewall)?

They go on to say that, “the SERC audit team intends to utilize Shodan during the early stages of its CIP audits, performing a search for the audited entity’s internet domain name,” and that the audit team will use the entity-provided list of BES Cyber Assets (they actually call them Critical Cyber Assets, but I’m pretty sure they mean BCA) and that any matches will probably be included in the sample selection for CIP-005 and CIP-007. They finish by saying, “the audited entity will be offered the opportunity to provide evidence the device and its associated network(s) are appropriately configured.”

I think that it’s pretty good advice to, as a defender, consider using Shodan to see if you have anything hanging out on the public Internet. From a commonsense perspective, there may be some sense to an auditor using Shodan in order to find devices which are publicly accessible, and therefore giving them a little extra attention, or at least making sure they aren’t missed. Ideally, though, auditors won’t find any control systems on Shodan, and so the search won’t add much value to the audit.

Cyber Asset Lesson Learned Finalized

NERC’s Standards Committee (SC) approved the Lesson Learned (LL)document on CIP-002-5: BES Cyber Assets on January 21, 2015. This document was originally released for comment on September 9, 2015, and the final document is dated December 7, 2015. It was created by NERC’s V5 Transition Advisory Group (V5TAG).

One of my criticisms of the LL process is that the LLs will be commented on, get revised, and then get approved and posted by the SC without industry having a chance to see or vet the revisions before they are finalized. While NERC has re-posted some of the LLs for comments, they’ve generally only done that when there are major revisions, and they definitely haven’t done so for all (or even most) of the Lessons Learned documents. This leads to a limited number of people actually having input into what the final, official, document contains. I believe this process has the potential to introduce flaws into the documents, and it happened again with the Cyber Asset LL.

The final document will be available from the “Related Documents” section of the CIP-002-5.1 page at NERC’s website, and the consideration of stakeholder comments will be available from the transition program page. The redlined version of the document is a little harder to find, but it is available from the SC agenda package for their January 21, 2016 meeting.

Sidenote: At one point, the redlined versions were not provided by the V5TAG or the SC. As part of my day  job I actually went and created a redlined version from the previous draft and the finalized draft for a couple of the LLs. That was a pain-in-the-butt, so the inclusion of the redlined versions the last few times finalized LLs were released was much appreciated.

Anyways, the purpose of this post is to look at some of what is in the LL and some of what was taken out of it.

When Cyber Assets meet a threshold of BES impact they become BES Cyber Assets (BCA) which may be grouped by responsible entities into BES Cyber Systems (BCS)

While I don’t disagree with this statement, I think it may be misleading to some people, specifically the word “threshold”. If the Cyber Asset has an adverse impact on the BES within 15 minutes of being misused, degraded, or unavailable, then it is a BES Cyber Asset.[^This is a simplified definition, obviously, and there are more qualifiers and requirements, but this summary works for now.] While the 15 minutes could be considered a threshold, there is no allowance in the standards for the adverse impact to have some minimum threshold. The way the sentence in the LL is worded, with “threshold” directly preceding “of BES impact” could cause some confusion for somebody who associates “BES impact” with the “adverse impact” from the BES Cyber Asset definition.

Some study participants assessed each functional system at a site or facility to determine its potential to adversely impact the BES in 15 minutes or less…Other study participants identified all Cyber Assets, grouped them into BES Cyber Systems, and evaluated the impact of the resulting system.

I like how this provides two approaches to achieving the goal of identifying BES Cyber Assets. This is a good example of an LL being written in such a way that it acknowledges that, “there may be other legitimate ways to fulfill the obligations of the requirements that are not expressed within this supporting document,” as the LL puts it in the introductory paragraph.

Did the device’s function directly impact the reliable operation of a BES asset?

The word “directly” was removed from here, and there were similar edits at other places. This was done in response to a comment from ACES, and seems like a good decision, although it leads to a problem discussed next. There’s nothing in the definition of BES Cyber Asset that talks about a difference between “direct” and “indirect” impacts, and so adding that concept here seems like it would add to confusion insteading of making things more clear.

Did the device function as an EACMS, PACS, or Intermediate System? These types of devices, such as firewalls, intrusion prevention systems or physical access controllers and others that perform security functions may indeed have an adverse impact if they are unavailable or misused to allow unauthorized access or deny authorized electronic or physical access, but it is an indirect impact. These devices have their own definitions and requirements in the CIP version 5 Reliability Standards and therefore are not considered BCAs due solely to their impact.

This is the most glaring example in this document of the problem of not having LLs vetted by industry after being edited. I do not think these changes would have made it through a public vetting process, if there had been one. In the first part of this paragraph, they are saying straight-up that firewalls, IPSs, or other security devices can have an adverse iimpact if they are unavailable or misused. While these devices do have their own definitions, there is nothing in the definition of BES Cyber Asset that would imply that an entire class of devices can be categorically excluded from the requirement to determine whether they are BES Cyber Assets. By removing those last five words from this paragraph in the draft LL, the V5TAG has gone beyond providing an example of a compliant approach, and has gone into the terrority of interpreting the standards.

There are programmable devices that may have an impact, but have no way to change their executing code, have no concept of a user or authentication, have no ports/services, have no network connectivity, no concept of event logs or alerting, no patches or updates, andor are located in areas where physical security perimeters cannot be established. For these devices, the ERO will allow (for the purposes of assessing compliance with the standards) Responsible Entities to only protect those devices to the extent capable.

The last sentence in this excerpt had, by far, the most comments related to it. EEI went so far as to say that unless the sentence was removed they couldn’t support the document being submitted to the SC for approval. The problem with the sentence was that it implies that non-programmable devices, if they would have an adverse impact within 15 minutes of their misuse, degradation, or unavailability, would have to be protected to the extent capable. The definition of Cyber Asset, though, explicitly states that a Cyber Asset is programmable. That means that a number of new devices would now be in-scope for the CIP standards.

They changed this table to say they were identifying Cyber Assets typically “evaluated” as BES Cyber Assets instead of saying they were typically “identified” as BCA. This change seems to be more in line with what the purpose of the list of device types originally was.

The transition study participants found a lack of clarity in the CIPV5 Reliability Standards because they do not specifically define the “programmable electronic device” component of the BES Cyber Asset NERC Glossary term. Consequently, the CIP V5 Transition Advisory Group referred the identified issues to be evaluated for standards development.

This was added to the document. While a better definitio of Cyber Asset or a definition of “programmable” is needed, unfortunately the creation of an SDT will not be a very quick process, so that clarification likely won’t help most entities very much in the near- to mid-term future.

Another Example of Why Prescriptive Regulations Fail

Last week, the Department of Homeland Security’s ICS-CERT released an advisory for the Advantech WebAccess product. The advisory detailed 15 different vulnerabilities, reported by at least seven different security researchers. They said that the vulnerabilities could be exploited remotely, and that an attacker with “low skill” would be able to eploit the vulnerabilities.

Patrick Coyle pointed out that the release notes and press release for this update, released by Advantech in late December, don’t have any indication that the update has any security-related patches. This, despite having 15 different vulnerabilities being fixed!

For people who are lucky enough to live in the NERC CIP world, this creates an interesting situation. Organizations are required to monitor a “patch source” for cyber security patches. This is likely to be the vendor, in this case Advantech. (Note: I don’t want it to sound like I’m picking on Advantech. The same situation happens with other vendors, too.) If the patch isn’t security-related, then the organization is not required to apply the patch. If the patch is security-related, then they either need to apply the patch or else create a mitigation plan for how they will prevent the vulnerability from being exploited.

Since the Advantech release notes didn’t say anything about this patch being security-related, the organization would have stopped there. Now, a month later, ICS-CERT releases their advisory and we found out, “Oh, that patch was security-related.” Here’s the kicker, though: The organization isn’t required to scour the Internet looking for vulnerability reports. They aren’t required to monitor ICS-CERT, they’re not required to monitor Full Disclosure, nothing. They did what they were required to do when they checked Advantech’s release notes and didn’t find anything that said it was security-related.

So now, we’re in a situation where the organization has a system with at least 15 publicly disclosed vulnerabilities. For the CIP patch management process, they did what they were supposed to do. The compliance people checked to make sure that they evaluated the Advantech patch. The operations people have a product they’re using, and if it’s working, they aren’t going to want to apply an update. The security people may or may not exist, and if they do exist, they may or may not be monitoring ICS-CERT (although I hope they would be).

The bottom line is, you have an entity which is fully compliant with the NERC CIP regulations, which have regulations in place requiring you to apply security patches, yet this security patch hasn’t been applied.

Automatically Open All Reading List Items

As part of my job, I write a weekly newsletter, and as part of that I need to keep up with the current news, read articles, and write summaries of them. I use daily emails from a few sources (such as the CyberWire and Politico’s Morning Cybersecurity) and a RSS reader to collect possible stories. A lot of time, I use my phone when I have a few spare moments and mark articles that might be worth reading more in-depth, and I use Safari’s Reading List to do that.Reading List has some problems, though. You can’t easily open every article in the Reading List. Since I generally clear it out every day, there might be anywhere from just a couple articles up to about 15-20 at the most. I wanted to be able to open every article and then clear the list, so that I can process the articles while I’m at work. Apple doesn’t provide a good way to do this, though, so I used a combination of AppleScript and Keyboard Maestro to automate it.

I chose AppleScript to open the articles. Apple doesn’t provide easy access to your Reading List. Instead, it’s saved as part of your Safari Bookmarks file. I used this post by Jacques Rioux as a template, and modified it a bit to just get the URLs of articles in the Reading List.

-- Sets tfile to look at Safari's bookmarks file, which contains the Reading List
set tfile to (path to library folder from user domain as text) & "Safari:bookmarks.plist"

-- Creats a blank list, which will later have Reading List items added to it.
set theURLs to {}

tell application "System Events"

    -- Check each item in the Safari bookmarks file 
    repeat with i in (property list items of property list item "Children" of property list file tfile)
        tell i to try

            -- Check the item to see if it is a part of the Reading List
            if value of property list item "Title" = "com.apple.ReadingList" then
                repeat with thisDict in (get value of property list item "Children")

                    -- Add item to list of URLs to open
                    tell thisDict to set end of theURLs to its URLString
                end repeat
                exit repeat
            end if
        end try
    end repeat
end tell

This created a list of URLs from my Reading List. Then, I used some more AppleScript to take that list of URLs and open them in Safari. First, this checks if the Reading List is empty. If there’s no articles in it, then the script stops. Otherwise, I take the first URL and open it in a new window, then I take each subsequent URL and open it in a new tab. This section of the script is based off of this answer on Stack Overflow.

tell application "Safari"
    -- If there are no items in the Reading List then do not open Safari
    -- TODO change this to pop up a notifcation that the Reading List is empty
    if theURLs = {} then
        return
    else

        -- Get first item of list so that it can be used to create a new window, while the rest of the list are used to create new tabs..
        set {firstURL, restURLs} to {item 1 of theURLs, rest of theURLs}
    end if

    -- Make new window with the first URL in the Reading List
    make new document at end of documents with properties {URL:firstURL}

    -- Make new tabs with each of the other URLs in the Reading List
    tell window 1
        repeat with theURL in restURLs
            make new tab at end of tabs with properties {URL:theURL}
        end repeat
    end tell
end tell

I then used Keyboard Maestro in order to run this script. This let me launch it from a hotkey shortcut, or what I’ve been doing more often, using the Keyboard Maestro menubar icon to run it. After the AppleScript is executed, then I activate Safari and use Keyboard Maestro’s ability to choose menu items to open a new tab.

Since my new tabs open with the Reading List visible in the side bar, then I can just right-click an article on the sidebar and clear my Reading List.

In the future, I’d like to automate the clearing of items in the Reading List. Apple doesn’t provide an easy way to do so programmatically, and so I think I might have to do it by using Keyboard Maestro’s GUI scripting. I don’t like to do that because the scripts aren’t very resilient, and often break when anything is updated. I’ll have to see how annoyed I am by having to right-click in order to clear the Reading List, and my level of annoyance at that will determine whether I put the time in to try to automate that part of the process.