Coventry Building Society Grid Card

Coventry Building Society have recently introduced the Grid Card as a simple form of 2-factor authentication. It replaces memorable words in the login process. Now the idea is that you require something you know (i.e. your password) and something you have (i.e. the Grid Card) to log in - 2 things = 2 factors. For more about authentication see this post.

How does it work? Very simply is the answer. During the log in process, you will be asked to enter the digits at 3 co-ordinates. For example: c3, d2 and j5 would mean that you enter 5, 6 and 3 (this is the example Coventry give). Is this better than a secret word? Yes, is the short answer. How many people will choose a memorable word that someone close to them could guess? Remember, that this isn't a password as such, it is expected to be a word and a word that means something to the user. The problem is that users cannot remember lots of passwords, so remembering two would be difficult. Also, having two passwords isn't really any different from having a longer, stronger password, it's still single-factor.

The idea behind the Grid Card is that you have a set of random numbers shared between you and the bank that are very hard to guess. I only say very hard to guess because I don't know how they generate the cards in the first place and if this isn't truly random - which it almost certainly won't be - then you can predict parts of the grid given other parts of it. Randomness is a rare but essential commodity. There are 50 co-ordinates on the card and Coventry ask for 3 each time, giving 19,600 possible combinations, assuming they'll never ask for the same co-ordinate more than once per login (order doesn't matter as we're told which grid squares). Does this mean that someone would have to log all 19,600 combinations before they could regenerate the card? No. Each co-ordinate appears 1,176 times in the 19,600. Each pair of co-ordinates appears 48 times. There are only really 17 unique combinations of co-ordinates such that they aren't repeated (and that's a cheat, because one co-ordinate will appear twice if we have 17 as 17x3=51). However, it is unlikely that these 17 would get asked for in succession, so it would take significantly more observations before we have the whole grid, but we won't need the whole grid before we're very likely to be able to login. Indeed, there's a 17.3% chance that at least one co-ordinate will be repeated on the next login. Also, a shoulder surfer with camera phone (or CCTV cameras) could take a photo of the whole card in one go, so this is an authentication mechanism to be used only in the 'safety' of your own home.

This is, however, a step in the right direction, so they should be commended for it. What else do you need to login to Coventry? Well, a Web ID and date of birth, both of which are easily pharmed. So the security is based solely on the password and Grid Card, which is better than two passwords. They do also have an anti-phishing technique bundled in there as well. When you sign up you choose a picture that they will display during your login along with your last login date and time. If the picture or date is incorrect then this isn't Coventry (or your account has been compromised). It's good to add a picture here, because many people don't actually check the last login date and time even if it's put up on the screen. The picture is obvious and hard to miss though. These mechanisms don't really stop spear phishing (or targeted phishing), but they do stop blanket or mass phishing attacks.

It's about time more banks started issuing 2-factor authentication for login and Coventry should be congratulated on being amongst the first. However, we have to be careful about how it's implemented.

Keylogging Trusteer's Rapport

Let's get some perspective on this first: no security product is 100% secure and just because there may be an obscure way round a product doesn't mean you shouldn't use it and that it won't protect you against a lot of attacks. How secure is your Anti-Virus (AV) product? Certainly not 100%, so we need layers of security. Rapport is another layer of security and could help protect your machine.

I have said in my previous post about this issue how well Trusteer dealt with me. So, now to the method of keylogging Trusteer. It's quite simple really, but requires a special setup. Rapport hooks onto the keyboard driver to prevent keylogging. However, if you invoke the remote desktop feature in Windows then a different keyboard driver is invoked, which Rapport cannot hook onto. So, if you're using a remote desktop connection into your machine then Rapport will not be giving you the full protection (it still has other layers of protection that work in this scenario).

Is this such a special case that you don't need to worry about it? Well not necessarily. There are a plethora of remote access software solutions available to users who are increasingly using them to access their machines at home or at work. There is also another technology that can be leveraged to cause this effect whilst the user is at the actual machine. Microsoft have introduced RemoteApps to the Windows desktop environment to allow for legacy applications to appear to run seamlessly on Windows 7. This is done via Virtual PC running another OS and the RAIL QFE update to allow applications to be exposed from a desktop machine as RemoteApps. However, we can use this technique to look back at the machine and expose the web browser as a RemoteApp, which the user should not notice.



As I say, it's a special case and not one a user would normally encounter, but it is possible. There are other issues with Trusteer as well, being able to capture the screen of protected websites and information leakage as highlighted on ReviewMyLife.co.uk here. It doesn't mean you shouldn't use Rapport though, just know and trust the machine that you're using. Basically, don't ever connect to any secure site or service from an untrusted machine, no matter what's installed on it.

Trusteer's Response to Issues with Rapport

I have been getting a lot of hits on this blog relating to Trusteer's Rapport, so I thought I would take a better look at the product. During my investigations, I was able to log keystrokes on a Windows 7 machine whilst accessing NatWest. However, the cause is as yet unknown as Rapport should be secure against this keylogger, so I'm not going to share the details here yet (there will be a video once Trusteer are happy there is no further threat).

I have had quite a dialogue with Trusteer over this potential problem and can report that their guys are pretty switched on, they picked up on this very quickly and are taking it extremely seriously. They are also realistic about all security products and have many layers of security in place within their own product. No security product is 100% secure - it can't be. The best measure of a product, in my opinion, is the company's response to potential problems. I have to admit that Trusteer have been exemplary here.

Why do I keep saying it's a potential problem when I have logged keystrokes? Well, under normal operating conditions this isn't possible with the keylogger used. Most home users won't have a machine set up like the test machine in this case.

Trusteer have also pointed out that keyloggers are not the main threat facing the banks at the moment and are of less use now than in the past. Rapport has several layers of security protecting the machine beyond keyloggers and blocking screen capture. One of he major plus points about Rapport is their anti-phishing and anti-pharming technologies. Although, again, these aren't perfect, it's better than nothing.

I don't agree totally with Trusteer here though. The problem with being able to log typed characters comes back to weak passwords and single-factor authentication. In this case, NatWest seem to require a customer ID, consisting of the user's date of birth and a 4 digit ID in the format ddmmyyxxxx, a 4 digit PIN and only a short password. Now, they will let any Customer ID through in this format whether it's valid or not (good from a security point of view as you don't know if you've got a valid Customer ID or not). However, clearly they allow 6 character passwords and then ask for three of them. So with one capture I can have 3 out of 4 PIN digits and half the password. We know people choose weak passwords that can be guessed. This becomes a crossword puzzle to make a 6 character password given three known characters. I would agree with Trusteer that keyloggers and screen capture shouldn't be a problem now, but it still is, as the banks cling onto simple username and password authentication, often with poor password policies.

If the banks move to 2-factor authentication and one-time passwords then most of this would be redundant, and Trusteer could concentrate on pushing us off to the correct site to avoid phishing and pharming attacks. Of course, these will become even more prevelant and sophisticated. Technology can't stop this alone, it has to be coupled with user education. Screen capture can still cause problems with strong authentication solutions, such as those using images or on-screen grids to generate one-time passwords.

So, what's the bottom line? Since my earlier posts, Rapport has come a long way with compatibility, etc. The tone of the marketing has also changed for the better and is more realistic (although some of the 44 partner banks could be doing more). So Rapport could be an additional layer of security to protect you, but you will still have to be vigilant. You must have an up-to-date, legitimate anti-virus/anti-malware product, firewall protection, tight controls on your browser and a cautious and skeptical approach to all communiations and links. Without these, Rapport isn't going to help you anyway.

Edit: video in later post - Keylogging Trusteer's Rapport

Cisco TACACS+ Password Length

I have recently come up against a problem with using the 'new' wireless network at work. We are using Cisco kit and TACACS+ to interface onto Microsoft's AD in the back end. Technically, usernames should be able to be up to 31 bytes long (not a problem there) and the password up to 254 bytes. However, the web portal implementation that we are running has a problem with my password. It would appear that passwords of up to 16 characters are fine, but passwords in excess of 16 characters don't work.

We are currently investigating this, as it seems like a real problem, especially as we are recommending that people switch to using longer pass phrases, in excess of 16 characters. Hopefully vendors will catch up with this soon, as many still have problems with so-called 'special characters' such as punctuation and other common symbols.

Cookieless Browser Tracking

We all know about tracking cookies and privacy. However, according to EFF it isn't necessary to use cookies to do a fair job of tracking your browser activities. According to their research browsers give 10.5 bits of identifying information in the userAgent string, which is supplied to the web server with every request. This is around a third of the information required to uniquely identify you.

They have set up a website to gather more data and give you a 'uniqueness' indicator for your browser, which you can find here. This data set is growing quite rapidly and will tell you how many of the userAgent strings they have received that are the same as yours. I managed to find a machine to test that was unique amongst the 195,000 machines they have tested. This means that someone could potentially track that machine even if cookies are disabled. Even if you come out with the same userAgent string as others, you can be narrowed down by using geolocation of your IP, browser plugins, installed fonts, screen resolution, etc. This isn't a new idea and others have tried it, like browserrecon. Of course if you have a static IP address then you are fairly easy to track anyway.

Various suggestions are made to help protect yourself, such as don't allow scripts to run on untrusted websites, which is fairly obvious. However, although this may reduce the amount of data given out from highs of 15.5 bits on a Blackberry or 15.3 bits on Debian, this won't stop the whole problem. It seems like the worst devices for giving out identifying information are Blackberry and Android phones, with minimum figures of over 12 bits. The best combination would seem to be FireFox running on Windows, which can be controlled down to only 4.6 bits (although highs are around double this), but this could just be because it's the most common combination.

What can you do? Don't visit untrusted sites. Also, you could change your userAgent string. It is just a text string stating the capabilities of your machine so that the web server can customise content to suit you. However, there is no real harm in tweaking this to fall in line with more common strings so that you are harder to track. You have to be careful here, because just removing most of the information will probably make your userAgent string unique. Alternatively, you could regularly change the string. Perhaps browsers should change the string with every connection? Plugins could do this, like User Agent Switcher. This would allow you to use different strings across different sites. Maybe hiding certain activities by temporarily switching the userAgent string would be useful.

FireFox and Opera are both quite easy to configure - type about:config or opera:config in the address bar respectively and navigate to the userAgent options. Internet Explorer is slightly more trickey, in that you have to make a registry change to alter the userAgent string. Navigate to [HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Internet Settings\5.0\User Agent] in regedit. Here you can create string values for 'Compatible', 'Version' and 'Platform' to control what is sent. Under the 'Post Platform' key are a whole bunch of additional parameters that will be added to the string, so you can change or remove these.

How secure is your AV Product?

We all use (or at least we should all use) an Anti-Virus (AV) product on our computer to protect it from malware (yes, that includes you Mac and Linux users as well). Rogue Anti-Malware is on the increase and users should be wary of what they install, but if we do choose a big vendor and pay money for it, does it protect our machine from all threats?

Well the answer is no. No security product can be 100% secure, but how secure are they actually? There have been a number of recent surveys and their results show that things are probably improving, but there's still a significant gap. AV-Comparatives.org showed that in their tests, G Data was the best with a 99.8% detection rate of known malware, with Norman being the worst of the 16 at 84.8%. Known malware was taken to be malware from a period of one year that ended 8 months prior to the test. This is important to stress; these weren't new malware instances, these were old known malware that all vendors will have seen and had time to develop their product to combat.

There's another potential issue as well. What settings do you use on your AV product? Do you use the default settings? Several products do come with the highest protection set as default, but not all. Kaspersky, Symantec and Sophos, for example, don't have the highest security settings by default (although Sophos, to their credit, asked AV-Comparatives to test them with default settings, unlike the other two who asked to be tested with settings changed to high security). McAfee use a cloud-based technology called Artemis, which is on by default, but requires an internet connection. Their test scores come down from 98.7% detection rate when online to 92.6% when offline. So be wary about the settings that you use and the mode of use as well, as it can make a big difference.

AV-Test.org also performed similar tests with more current malware, with similar results. In their tests, Symantec came out top with a score of 98% malware detected and Trend Micro with 83.3%. I'll pick out a few big names so that I can give you average figures from both testing labs.

Detection & Blocking Rates for some Major AV Products

Product	Existing Detection	Blocking	Live Detection
Symantec	98.2%	92.8%	35.5%
Kaspersky	96.1%	89.9%	41.0%
McAfee	93.0%	86.7%	45.5%
AVG	93.1%	84.2%	40.0%
F-Secure	91.9%	80.2%	42.0%

This isn't the full story though. The above tests are detected existing malware. There are two other metrics that we need to look at. The first of these is the removal or blocking rate. This is the percentage of malware instances that were blocked or removed by the AV product. The others will have infected the machine. AV-Test.org correctly point out that this is a much more important metric than detected malware, as if an AV product detects it but still allows it to install, then you are only marginally better off than if you didn't know about it at all - your machine is still infected. Their tests show that the blocking rates are a chunk down from the detection rates, with the best now being PC Tools at 94.8% and the worst being CA Internet Security at 73.5%. Blocking rate figures for the set of AV products are also given in the table above.

The final thing to consider is the detection rate of new malware that hasn't been seen before, i.e. from live attacks. Cyveillance performed a set of tests sending live attack malware through a set of the top AV products on a daily basis to see how they performed. In their tests, cloud-based McAfee came out top at 44% and VirusBuster bottom on 16%. AV-Comparatives performed a similar test and came out with slightly better results, ranging from AVIRA on 74% down to Norman on 32%. Again, I have averaged their figures to include in the table above.

Conclusion: you could use more than one AV product as long as they don't conflict. However, it is essential that you keep the product up-to-date at all times and configure it for maximum protection.

Pragmatic Approach to Security

When dealing with security, we must be pragmatic. The resources that an organisation can dedicate to security are limited in terms of time, staff, budget, expertise, etc. Also, perfectly secure systems do not exist - accidents, attacks and penetrations will happen in the end, so plan to deal with them at the outset. Recovery after a breach must be just as much of the planning as the mitigation of the breach in the first place. We all insure our cars, hoping never to call on it, and then try desperately to avoid having any accidents, getting the car stolen or vandalized. However, in the end, a lot of us will end up claiming on the insurance at some point, no matter how careful we are. The same is true of security.



We have to see the bigger picture and align the use of resources with the company's mission. There comes a point when a small amount more security costs a lot more money, time, management effort and is much less user-friendly. Wouldn't it impact the business less if we take the hit and recover quickly and smoothly? Often the answer is yes. We have to find the optimal solution for that particular organisation. The graph above shows that as we increase the security of our system the cost associated with breaches of security comes down, as we have fewer breaches. However, this cost will never be zero, as we will always have breaches. Indeed, breaches may still cost a lot of money but, hopefully they will be few and far between. Conversely, as our security increases, the cost of our countermeasures goes up. Therefore, the total cost will decrease with more security initially, then increase again as the countermeasures become increasingly expensive for less and less improvement to security.

These curves and the overall graph will be different for each organisation. The point I'm trying to make is that we should accept that there is no perfect security, do the best job we can, given the resources allocated, and plan for how we will recover from any breaches in security, be they minor or major. The problem comes when deciding what assets should be given priority and what is the best allocation of resources for a specific organisation. This is where security risk assessments come in. For more about security assessments and risks, see my previous post.