Blog posts by Martin Kuppinger
A few days ago, Microsoft announced Azure Confidential Computing. As the name implies, the technology is about adding a new layer of protection to cloud services, specifically Microsoft Azure, but also Windows 10 and Windows Server 2016 running in other public cloud infrastructures on specific hardware.
The foundation for Azure Confidential Computing are so-called TEEs (Trusted Execution Environments). Such environments protect the code running in that environment and data used by the code from other parties’ access. Neither administrators, neither people having direct access to hardware, nor attackers that gain access to administrator accounts can bypass that protection layer – at least this is what the TEE concept promises.
Based on TEEs, data can be held encrypted in the cloud services and their data stores and are only decrypted and processed within the TEE. That means that data is not always encrypted, but it remains – if the application is implemented correctly – encrypted in the accessible areas of the public cloud.
For now, there are two supported TEEs. One is Virtual Secure Mode, a software-based TEE that is based on Microsoft Hyper-V in Windows 10 and Windows Server 2016. The other is Intel SGX (Software Guard Extensions), which is a hardware-based TEE. Based on Intel SGX, secure TEEs can be used outside of the Microsoft Azure Cloud.
Microsoft has been using such technologies as part of their Coco Framework for enterprise blockchain networks for some weeks already, and now is moving support to Microsoft SQL Server and Azure SQL Database. This is achieved by delegating computations on sensitive data to an “enclave”, which is based on a TEE. However, Azure Confidential Computing supports broader use of this capability for various types of data.
Microsoft Azure Confidential Computing, which is available in an early adopter’s program, is a great improvement for security and confidentiality in public cloud environments and will enable customers to port workloads to the cloud which, so far, have been considered as being too sensitive. The announcement stands in line with the recent IBM announcement for their IBM Z14 systems, where factually the entire system acts as a TEE. While the use of TEEs in Azure Confidential Computing is restricted to parts of the application that are moved to the TEE specifically, both announcements are about significantly increasing the level of security in computing. That is good news.
As a long-term observer of the IAM market, KuppingerCole finds it interesting to see the change in both the size of investments and the type of investors in this market. Just recently, ForgeRock announced an $88 million round in series D funding. This follows other major investments in IAM vendors such as Okta, Ping Identity, and SailPoint, to name a few.
What is interesting with the recent funding for ForgeRock is that KKR appears on the list, one of the very big names amongst the investors. I found that particularly telling because it means that IAM is now on the radar of a different type of investor, beyond the more focused IT- and Information Security-focused investors we have primarily seen until now.
Obviously, such major investment helps ForgeRock to continue their growth, to further expand their product offerings, and strengthens their market position. We will closely follow the plans and releases of ForgeRock for their Identity Platform and keep you up-to-date.
A couple of days ago, DIACC (Digital ID & Authentication Council of Canada) together with IBM Canada and the Province of British Columbia released information about a PoC (Proof of Concept) for moving corporate registrations to a blockchain-based register. The PoC, which used the Hyperledger Fabric, was for both corporate registries of a single province and across multiple jurisdictions.
Such registries, be it corporate registries, land register, or other types of decentralized ledgers, are the sweet spot for blockchains. Registration is decentralized. The registries and ledgers must be tamper-resistant. The data must be sequenced and, in many cases, time-stamped. All these are the fundamental characteristics of blockchains. Simply said: It is far easier to construct solutions for such registries and ledgers based on blockchain technology than it is based on traditional technologies such as relational databases.
Here, the use case and the solution fit perfectly well. Given that improving such registries and other ledgers can have a massive economic and societal impact, it also proves the value of blockchains, well beyond cryptocurrencies. Blockchains are here to stay, and I expect to see a strong uptake of use cases around registers and ledgers in the next couple of years – beyond PoCs towards practical, widely deployed and used implementations. Honestly, every investment in IT solutions for such registers etc. should be revisited, evaluating whether stopping it and restarting based on blockchains isn’t the better choice nowadays. In most cases it will turn out that blockchains are the better choice.
In a recent press release, IBM announced that they are moving security to a new level, with “pervasively encrypted data, all the time at any scale”. That sounded cool and, after talks with IBM, I must admit that it is cool. However, it is “only” on their IBM Z mainframe system, specifically the IBM Z14.
By massively increasing the encryption capabilities on the processor and through a system architecture that is designed from scratch to meet the highest security requirements, these systems can hold data encrypted at any time, with IBM claiming support of up to 12 billion encrypted transactions per day. Business data and applications running on IBM Z can be better protected than ever before – and better than on any other platform.
One could argue that this is happening in a system environment that is slowly dying. However, IBM in fact has won a significant number of new customers in the past year. Furthermore, while this is targeted as of now at mainframe customers, there is already one service that is accessible via the cloud: a blockchain service where secure containers for the blockchain are operated in the IBM Cloud in various datacenters across the globe.
It will be interesting to see whether and when IBM will make more of these pervasive encryption capabilities available as a cloud service or in other forms for organizations not running their own mainframes. The big challenge here obviously will be end-to-end security. If there is a highly secure mainframe-based backend environment, but applications accessing these services through secure APIs from less secure frontend environments, there will remain a security risk. Unfortunately, other platforms don’t come with the same level of built-in security and encryption power as the new IBM Z mainframe.
Such a gap between what is available (or will be available soon) on the mainframe and what we find on other platforms is not new. Virtualization was available on the mainframe way before the age of VMware and other PC virtualization software started. Systems for dynamically authorizing requests at runtime such as RACF are the norm in mainframe environments, while the same approach and standards such as XACML are still struggling in the non-mainframe world.
With its new announcement, IBM on one hand again shows that many interesting capabilities are introduced on mainframes first, while also demonstrating a potential path into the future of mainframes: as the system that manages the highest security environments and maybe in future acts as the secure backend environment, accessible via the cloud. I’d love to see the latter.
Today, Adobe announced that Flash will go end-of-life. Without any doubt, this is great news from an Information Security perspective. Adobe Flash counted for a significant portion of the most severe exploits as, among others, F-Secure has analyzed. I also wrote about this topic back in 2012 in this blog.
From my perspective, and as stated in my post from 2012, the biggest challenge hasn’t been the number of vulnerabilities as such, but the combination of vulnerabilities with the inability to fix them quickly and the lack a well-working patch management approach.
With the shift to standards such as HTML5, today’s announcement finally moves Adobe Flash into the state of a “living zombie” – and with vendors such as Apple and Microsoft either not supporting it or limiting its use, we are ready to switch to better alternatives. Notably, the effective end-of-life date is the end of 2020, and it will still be in use after that. But there will be an end.
Clearly, there are and will be other vulnerabilities in other operating systems, browsers, applications, and so on. They will not go away. But one of the worst tools ever from a security perspective is finally reaching its demise. That is good, and it makes today a great day for Information Security.
In a response to the EC Commission, the EBA (European Banking Authority) rejected amendments on screen scraping in the PSD2 regulation (Revised Payment Services Directive) that had been pushed by several FInTechs. While it is still the Commission’s place to make the final decision, the statement of the EBA is clear. I fully support the position of the EBA: Screen scraping should be banned in future.
In a “manifesto”, 72 FinTechs had responded to the PSD2 RTS (Regulatory Technical Standards), focusing on the ban of screen scraping or as they named it, “direct access”. In other comments from that FinTech lobby, we can find statements such as “… sharing login details … is perfectly secure”. Nothing could be more wrong. Sharing login details with whomever never ever is perfectly secure.
Screen scraping involves sharing credentials and providing full access to financing services such as online banking to the FinTechs using these technologies. This concept is not new. It is widely used in such FinTech services because there has been a gap in APIs until now. PSD2 will change that, even while we might not end with a standardized API as quickly as we should.
But what is the reasoning of the FinTechs in insisting on screen scraping? The main arguments are that screen scraping is well-established and works well – and that it is secure. The latter obviously is wrong – neither sharing credentials nor injecting credentials into websites can earnestly be considered a secure approach. The other argument, screen scraping being something that works well, also is fundamentally wrong. Screen scraping relies on the target website or application to always have the same structure. Once it changes, the applications (in that case FinTech) accessing these services and websites must be changed as well. Such changes on the target systems might happen without prior notice.
I see two other arguments that FinTech lobby does not raise. One is about liability issues. If a customer gives his credentials to someone else, this is a fundamentally different situation regarding liability then in the structured access via APIs. Just read the terms and conditions of your bank regarding online banking.
The other argument is about limitations. PSD2 request providing APIs for AISP (Account Information Service Providers) and PISP (Payment Initiation Service Providers) – but only for these services. Thus, APIs might be more restrictive than screen scraping.
However, the EBA has very good arguments in favor of getting rid of screen scraping. One of the main targets of PSD2 is a better protection of customers in using online services. That is best achieved by a well-thought-out combination of SCA (Strong Customer Authentication) and defined, limited interfaces for TPPs (Third Party Providers) such as the FinTechs.
Clearly, this means a change for both the technical implementations of FinTech services that rely on screen scraping as, potentially, for the business models and the capabilities provided by these services. When looking at technical implementations, even while there is not yet an established standard API supported by all players, working with APIs is straightforward and far simpler than screen scraping ever can be. If there is not a standard API, work with a layered approach which maps your own, FinTech-internal, API layer to the various variants of the banks out there. There will not be that many variants, because the AISP and PISP services are defined.
Authentication and authorization can be done far better and more convenient to the customer if they are implemented from a customer perspective – I just recently wrote about this.
Yes, that means changes and even restrictions for the FinTechs. But there are good reasons for doing so. EBA is right on their position on screen scraping and hopefully, the EC Commission will finally share the EBA view.
The Revised Payment Services Directive (PSD2), an upcoming EC regulation, will have a massive impact on the Finance Industry. While the changes to the business are primarily based on the newly introduced TPPs (Third Party Providers), which can initiate payments and request access to account information, the rules for strong customer authentication (SCA) are tightened. The target is better protection for customers of financial online services.
Aside from a couple of exemptions such as small transactions below 30 € and the use of non-supervised payment machines, e.g. in parking lots, the basic rule is that 2FA (Two Factor Authentication) becomes mandatory. Under certain circumstances, 1FA combined with RBA (Risk Based Authentication) will continue to be allowed. I have explained various terms in an earlier post.
For the scenarios where 2FA is required, the obvious question is how best to do that. When looking at how banks and other services implemented 2FA (and 1FA) up until now, there is plenty of room for improvement. While many services, such as PayPal, still only mandate 1FA, there generally is little choice in which 2FA approach to use. Most banks mandate the use of one specific form of 2FA, e.g. relying on out-of-band SMS or a certain type of token.
However, PSD2 will change the play for financial institutions. It will open the fight for the customer: Who will provide the interface to the customer, who will directly interact with the customer? To win in that fight between traditional and new players, customer convenience is a key success factor. And customer convenience starts with the registration as a one-time action and continues with authentication as what the customers must do every time they access the service.
Until now, strong (and not so strong) authentication to financial services seems to have been driven by an inside-out way of thinking. The institutions think about what works best for them: what fits into their infrastructure; what is the cheapest yet compliant approach? For customers, this means that they must use what their service provider offers to them.
But the world is changing. Many users have their devices of choice, many of these with some form of built-in strong authentication. They have their preferred ways of interacting with services. They also want to use a convenient method for authentication. And in the upcoming world with TPPs that can form the new interface, so there will be competition.
Thus, it is about time to think SCA outside-in, from the customer perspective. The obvious solution is to move to Adaptive Authentication, which allows the use of all (PSD2 compliant) forms of 2FA and leaves it to the choice of the customer which he prefers. There must be flexibility for the customer. The technology is available, with platforms that support many, many different types of authenticators and their combinations for 2FA, but also with standards such as the FIDO Alliance standards that provide interoperability with the ever-growing and ever-changing consumer devices in use.
There is room for being both compliant to the SCA requirements of PSD2 and convenient for the customer. But that requires a move to an outside-in thinking, starting with what the customers want – and these many customers never only want one single choice, they want a real choice. Adaptive Authentication thus is a key success factor for doing SCA right in the days of PSD2.
Recently, I stumbled about the first marketing campaigns of vendors claiming that they have a “GDPR compliant” application or SaaS offering. GDPR stands for General Data Protection Regulation and is the upcoming EC regulation in that field, which also has an extraterritorial effect, because it applies to every organization doing business with EU residents. Unfortunately, neither SaaS services nor software can be GDPR compliant.
GDPR is a regulation for organizations that regulates how to protect the individual’s PII (Personally Identifiable Information), which includes all data that could potentially be used to identify an individual. Thus, organizations must enforce GDPR compliance, which includes, e.g., implementing the new principles for user consent such as informed and unambiguous consent per purpose; the right to be forgotten; and many other requirements. GDPR also states that software which is used to handle PII must follow the principles of Security by Design (SbD) and Privacy by Design (PbD). Both are rather fuzzy principles, not being formally defined yet.
Thus, a software vendor or SaaS provider could state that he believes he is following the SbD and PbD principles. But that does not make him GDPR compliant. It just builds the foundation for a customer, enabling that organization becoming GDPR compliant. But to put it clearly: An organization dealing with PII can be GDPR compliant. A service provider that acts as “data processor” in the context of GDPR can be GDPR compliant (for its part of the system). But a software application or a SaaS service only can provide the foundation for others to become GDPR compliant. There just is no such thing as GDPR compliant software.
Vendor marketing departments would be well advised to use such terms carefully, because claiming to provide a GDPR compliant offering might make their customers think that they just need to install certain software or turn the key of a turnkey SaaS solution and they are done. Nothing could be more misleading. There is so much more to do for an organization to become GDPR compliant, starting from processes and contracts to using the software or SaaS service the right way. Understanding what GDPR really means to an organization is the first step. KuppingerCole has plenty of information on GDPR.
Don’t hesitate to contact KuppingerCole via firstname.lastname@example.org for our brand-new offering of a GDPR Readiness Assessment, which is a lean approach in understanding where your organization is in your journey towards GDPR compliance and which steps you have to take – beyond just choosing a tool.
WannaCry counts, without any doubt, amongst the most widely publicized cyber-attacks of the year, although this notoriety may not necessarily be fully justified. Still, it has affected hospitals, public transport, and car manufacturing, to name just a few of the examples that became public. In an earlier blog post, I was looking at the role government agencies play. Here I look at businesses.
Let’s look at the facts: The exploit has been known for a while. A patch for the current Windows systems has been out for months, and I’ve seen multiple warnings in the press indicating the urgent need to apply the appropriate patches.
Unfortunately, that didn’t help, as the number of affected organizations around the world has demonstrated. Were those warnings ignored? Or had security teams missed them? Was that ignorance? Lack of established processes? If they had older Windows versions in place in sensitive areas, why haven’t they replaced them earlier? I could ask many more of these questions. Unfortunately, there is only one answer to them: human failure. There is no excuse.
Somewhere in the organizations affected, someone – most likely several people – have failed. Either they’ve failed by not doing IT security right (and we are not talking about the most sophisticated attacks, but simply about having procedures in place to react to security alerts) or by lacking adequate risk management. Or by a lack of sufficient budgets for addressing the most severe security risks. Unfortunately, most organizations still tend to ignore or belittle the risks we are facing.
Yes, there is no 100% security. But we are all supposed to know how to strengthen our cyber-attack resilience by having right people, right processes, and right tools in place. The right people to deal with alerts and incidents. The right processes for both preparing for and reacting to breaches. And the right tools to detect, protect, respond, and recover. Yes, we have a massive skills gap that is not easy to close. But at least to a certain extent, MSSPs (Managed Security Service Providers) are addressing this problem.
Unfortunately, most organizations don’t have enterprise-wide GRC programs covering all risks including IT security risks, and most organizations don’t have the processes defined for an adequate handling of alerts and incidents – to say nothing about having a fully operational CDC (Cyber Defense Center). Having one is a must for large organizations and organizations in critical industries. Others should work with partners or at least have adequate procedures to recover quickly.
Many organizations still rely on a few isolated, old-fashioned IT security tools. Yes, modern tools cost money. But that is not even where the problem starts. It starts with understanding which tools really help mitigating which risks; with selecting the best combination of tools; with having a plan. Alas, I have seen way too few well-thought-out security blueprints so far. Creating such blueprints is not rocket science. It does not require a lot of time. Why are so many organizations lacking these? Having them would allow for targeted investments in security technology that helps, and also for understanding the organizational consequences. Just think about the intersection of IT security and patch management.
To react to security incidents quickly and efficiently, organizations need a CDC staffed with people, having defined processes in place for breach and incident response, and being well integrated into the overall Risk Management processes, as depicted in the picture below.
Such planning not only includes a formal structure of a CDC, but plans for handling emergencies, ensuring business continuity, and communication in cases of breaches. As there is no 100% security, there always will be remaining risks. No problem with that. But these must be known and there must be a plan in place to react in case of an incident.
Attacks like WannaCry pose a massive risk for organizations and their customers - or, in the case of healthcare, patients. This is a duty for the C-level – the CISOs, the CIOs, the CFOs, and the CEOs – to take finally responsibility and start planning for the next such attack in advance.
Just a few days ago, in my opening keynote at our European Identity & Cloud Conference I talked about the strong urge to move to more advanced security technologies, particularly cognitive security, to close the skill gap we observe in information security, but also to strengthen our resilience towards cyberattacks. The Friday after that keynote, as I was travelling back from the conference, reports about the massive attack caused by the “WannaCry” malware hit the news. A couple of days later, after the dust has settled, it is time for a few thoughts about the consequences. In this post, I look at the role government agencies play in increasing cyber-risks, while I’ll be looking at the mistakes enterprises have made in a separate post.
Let me start with publishing a figure I used in one of my slides. When looking at how attacks are executed, we can distinguish between five phases – which are more towards a “red hot” or a less critical “green” state. At the beginning, the attack is created. Then it starts spreading out and remains undetected for a while – sometimes only shortly, sometimes for years. This is the most critical phase, because the vulnerabilities used by the malware exist, but aren’t sufficiently protected. During that phase, the exploit is called a “zero-day exploit”, a somewhat misleading term because there might be many days until day zero when the exploit attacks. The term refers to the fact that attacks occur from day zero, the day when the vulnerability is detected and countermeasures can start. In earlier years, there was a belief that there are no attacks that start before a vulnerability is discovered – a naïve belief.
Here, phase 3 begins, with the detection of the exploit, analysis, and creation of countermeasures, most commonly hot fixes (that have been tested only a little and usually must be installed manually) and patches (better tested and cleared for automated deployment). From there on, during the phase 4, patches are distributed and applied.
Ideally, there would be no phase 5, but as we all know, many systems are not patched automatically or, for legacy operating systems, no patches are provided at all. This leaves a significant number of systems unpatched, such as in the case of WannaCry. Notably, there were alerts back in March that warned about that specific vulnerability and strongly recommended to patch potentially affected systems immediately.
In fact, for the first two phases we must deal with unknown attack patterns and assume that these exist, but we don’t know about them yet. This is a valid assumption, given that new exploits across all operating systems (yes, also for Linux, MacOS or Android) are detected regularly. Afterwards, the patterns are known and can be addressed.
In that phase, we can search for indicators of known attack patterns. Before we know these, we can only look for anomalies in behavior. But that is a separate topic, which has been hot at this year’s EIC.
So, why do I sometimes wanna cry about the irresponsibility and shortsightedness of government agencies? Because of what they do in phases 1 and 2. The NSA has been accused of having been aware of the exploit for quite a while, without notifying Microsoft and thus without allowing them to create a patch. Government organizations from all over the world know a lot about exploits without informing vendors about them. They even create backdoors to systems and software, which potentially can be used by attackers as well. While there are reasons for that (cyber-defense, running own nation-state attacks, counter-terrorism, etc.), there are also reasons against it. I don’t want to judge their behavior; however, it seems that many government agencies are not sufficiently aware of the consequences of creating their own malware for their purposes, not notifying vendors about exploits, and mandating backdoors in security products. I doubt that the agencies that do so can sufficiently balance their own specific interests with the cyber-risks they cause for the economies of their own and other countries.
There are some obvious risks. The first one is that a lack of notification extends the phase 2 and attacks stay undetected longer. It would be naïve to assume that only one government agency knows about an exploit. It might be well detected by other agencies, friends or enemies. It might also have been detected by cyber-criminals. This gets even worse when governments buy information about exploits from hackers to use it for their own purposes. It would be also naïve to believe that only that hacker has found/will find that exploit or that he will sell that information only once.
As a consequence, the entire economy is put at risk. People might die in hospitals because of such attacks, public transport might break down and so on. Critical infrastructures become more vulnerable.
Creating own malware and requesting backdoors bears the same risks. Malware will be detected sooner or later, and backdoors also might be opened by the bad guys, whoever they are. The former results in new malware that is created based on the initial one, with some modifications. The latter leads to additional vulnerabilities. The challenge is simply that, in contrast with physical attacks, there is little needed to create a new malware based on an existing attack pattern. Once detected, the knowledge about it is freely available and it just takes a decent software developer to create a new strain. In other words, by creating own malware, government agencies create and publicize blueprints for the next attacks – and sooner or later someone will discover such a blueprint and use it. Cyber-risk for all organizations is thus increased.
This is not a new finding. Many people including myself have been hinting about this dilemma for long in publications and presentations. While, being a dilemma, it is not easy to solve, we need at least to have an open debate on this and we need government agencies that work in this field to at least understand the consequences of what they are doing and balance this with the overall public interest. Not easy to do, but we do need to get away from government agencies acting irresponsibly and shortsighted.
Register now for KuppingerCole Select and get your free 30-day access to a great selection of KuppingerCole research materials and to live trainings.
The EU GDPR (General Data Protection Regulation), becoming effective May 25, 2018, will have a global impact not only on data privacy, but on the interaction between businesses and their customers and consumers. Organizations must not restrict their GDPR initiatives to technical changes in consent management or PII protection, but need to review how they onboard customers and consumers and how to convince these of giving consent, but also review the amount and purposes of PII they collect. The impact of GDPR on businesses will be far bigger than most currently expect. [...]