As we inch closer to the midpoint of 2015, we find ourselves in a drastically different world of both connectivity and security. Many of us switch devices throughout the day, from phone to tablet to laptop and back again. Even in corporate workplaces, the ubiquity of mobile devices has come to stay (in spite of the clamoring and frustration of many IT directors!). The efficiency and ease of use of integrated mobile and tethered devices propels many business solutions today. The various forms of cloud resources link all this together – whether personal or professional.
But this enormous change in topology has introduced very significant security implications, most of which are not really well dealt with using current tools, let alone software or devices that were ‘state of the art’ only a few years ago.
What does this mean for the user – whether personal or business? How do network admins and others that must protect their networks and systems deal with these new realities? That’s the focus of the brief discussion to follow.
No More Walls…
The pace of change in the ‘Internet’ is astounding. Even seasoned professionals who work and develop in this sector struggle to keep up. Every day when I read periodicals, news, research, feeds, etc. I discover something I didn’t know the day before. The ‘technosphere’ is actually expanding faster than our collective awareness – instead of hearing that such-and-such is being thought about, or hopefully will be invented in a few years, we are told that the app or hardware already exists and has a userbase of thousands!
One of the most fundamental changes in the last few years is the transition from ‘point-to-point’ connectivity to a ‘mesh’ connectivity. Even a single device, such as a phone or tablet, may be simultaneously connected to multiple clouds and applications – often in highly disparate geographical locations. The old tried-and-true methodology for securing servers, sessions and other IT functions was to ‘enclose’ the storage, servers and applications within one or more perimeters – then protect those ‘walled gardens’ with firewalls and other intrusion detection devices.
Now that we reach out every minute across boundaries to remotely hosted applications, storage and processes the very concept of perimeter protection is no longer valid nor functional.
Even the Washing Machine Needs Protection
Another big challenge for today’s security paradigm is the ever-growing “Internet of Things” (IoT). As more and more everyday devices become network-enabled, from thermostats to washing machines, door locks to on-shelf merchandise sensors – an entirely new set of security issues has been created. Already the M2M (Machine to Machine) communications are several orders of magnitude greater than sessions involving humans logging into machines.
This trend is set to literally explode over the next few years, with an estimated 50 billion devices being interconnected by 2020 (up from 8.7 billion in 2012). That’s a 6x increase in just 8 years… The real headache behind this (from a security point of view) is the amount of connections and sessions that each of these devices will generate. It doesn’t take much combinatorial math to see that literally trillions of simultaneous sessions will be occurring world-wide (and even in space… the ISS has recently completed upgrades to push 3Mbps channels to 300Mbps – a 100x increase in bandwidth – to support the massive data requirements of newer scientific experiments).
There is simply no way to put a ‘wall’ around this many sessions that are occurring in such a disparate manner. An entirely new paradigm is required to effectively secure and monitor data access and movement in this environment.
How Do You Make Bulletproof Spaghetti?
If you imagine the session connections from devices to other devices as strands of pasta in a boiling pot of water – constantly moving and changing in shape – and then wanted to encase each strand in an impermeable shield…. well you get the picture. There must be a better way… There are a number of efforts underway currently from different researchers, startups and vendors to address this situation – but there is no ‘magic bullet’ yet, nor is there even a complete consensus on what method may be best to solve this dilemma.
One way to attempt to resolve this need for secure computation is to break the problem down into the two main constituents: authentication of whom/what; and then protection of the “trust” that is given by the authentication. The first part (authentication) can be addressed with multiple-factor login methods: combinations of biometrics, one-time codes, previously registered ‘trusted devices’, etc. I’ve written on these issues here earlier. The second part: what does a person or machine have access to once authenticated – and how to protect those assets if the authentication is breached – is a much thornier problem.
In fact, from my perspective the best method involves a rather drastically different way of computing in the first place – one that would not have been possible only a few years ago. Essentially what I am suggesting is a fully virtualized environment where each session instance is ‘built’ for the duration of that session; only exposes the immediate assets required to complete the transactions associated with that session; and abstracts the ‘devices’ (whether they be humans or machines) from each other to the greatest degree possible.
While this may sound a bit complicated at first, the good news is that we are already moving in that direction, in terms of computational strategy. Most large scale cloud environments already use virtualization to a large degree, and the process of building up and tearing down virtual instances has become highly automated and very, very fast.
In addition, for some time now the industry has been moving towards thinner and more specific apps (such as found on phones and tablets) as opposed to massive thick client applications such as MS Office, SAP and other enterprise builds that fit far more readily into the old “protected perimeter” form of computing.
In addition (and I’m not making a point of picking on a particular vendor here, it’s just that this issue is a “fact of nature”) the Windows API model is just not secure any more. Due to the requirement of backwards compatibility – to a time where the security threats of today were not envisioned at all – many of the APIs are full of security holes. It’s a constant game of reactively patching vulnerabilities once discovered. This process cannot be sustained to support the level of future connectivity and distributed processing towards which we are moving.
Smaller, lightweight apps have fewer moving parts, and therefore by their very nature are easier to implement, virtualize, protect – and replace entirely should that be necessary. To use just an example: MS Word is a powerful ‘word processor’ – which has grown to integrate and support a rather vast range of capabilities including artwork, page layout, mailing list management/distribution, etc. etc. Every instance of this app includes all the functionality, of which 90% is unused (typically) during any one session instance.
If this “app” was broken down into many smaller “applets” that called on each other as required, and were made available to the user on the fly during the ‘session’ the entire compute environment becomes more dynamic, flexible and easier to protect.
Lowering the Threat Surface
One of the largest security challenges of a highly distributed compute environment – such as is presented by the typical hybrid cloud / world-wide / mobile device ecosystem that is rapidly becoming the norm – is the very large ‘threat surface’ that is exposed to potential hackers or other unauthorized access.
As more and more devices are interconnected – and data is interchanged and aggregated from millions of sensors, beacons and other new entities, the potential for breaches is increased exponentially. It is mathematically impossible to proactively secure every one of these connections – or even monitor them on an individual basis. Some new form of security paradigm is required that will, by its very nature, protect and inhibit breaches of the network.
Fortunately, we do have an excellent model on which to base this new type of security mechanism: the human immune system. The ‘threat surface’ of the human body is immense, when viewed at a cellular level. The number of pathogens that continually attempt to violate the human body systems are vastly greater than even the number of hackers and other malevolent entities in the IT world.
The conscious human brain could not even begin to attempt to monitor and react to every threat that the hordes of bacteria, viruses and other pathogens bring against the body ecosystem. About 99% of such defensive response mechanisms are ‘automatic’ and go unnoticed by our awareness. Only when things get ‘out of control’ and the symptoms tell us that the normal defense mechanisms need assistance do we notice things like a sore throat, an ache, or in more severe cases: bleeding or chest pain. We need a similar set of layered defense mechanisms that act completely automatically against threats to deal with the sheer numbers and variations of attack vectors that are becoming endemic in today’s new hyper-connected computational fabric.
A Two-Phased Approach to Hyper-Security
Our new hyper-connected reality requires an equally robust and all-encompassing security model: Hyper-Security. In principle, an approach that combines the absolute minimal exposure of any assets, applications or connectivity with a corresponding ‘shielding’ of the session using techniques to be discussed shortly can provide an extremely secure, scalable and efficient environment.
Phase One – building user ‘sessions’ (whether that user is a machine or a human) that expose the least possible amount of threat surface while providing all the functionality required during that session – has been touched on earlier during our discussion of virtualized compute environments. The big paradigm shift here is that security is ‘built in’ to the applications, data storage structures and communications interface at a molecular level. This is similar to how the human body systems are organized, which in addition to the actual immune systems and other proactive ‘security’ entities, help naturally limit any damage caused by pathogens.
This type of architecture simply cannot be ‘backed in’ to legacy OS systems – but it’s time that many of these are moved to the shelf anyway: they are becoming more and more clumsy in the face of highly virtualized environments, not to mention the extreme amount of time/cost to maintaining these outdated systems. Having some kind of attachment or allegiance to an OS today is as archaic as showing a preference for a Clydesdale vs a Palomino in the world of Ferraris and Teslas… Really all that matters today is the user experience, reliability and security. How something gets done should not matter any more, even to highly technical users, any more than knowing exactly which endocrines are secreted by our Islets of Langerhans (some small bits of the pancreas that produce some incredibly important things like insulin). These things must work (otherwise humans get diabetes or computers fail to process) but very few of us need to know the details.
Although the concept of this distributed, minimalistic and virtualized compute environment is simple, the details can become a bit complex – I’ll reserve further discussion for a future post.
To summarize, the security provided by this new architecture is one of prevention, limitation of damage and ease of applying proactive security measures (to be discussed next).
Phase Two – the protection of the compute sessions from either internal or external threat mechanisms – also requires a novel approach that is suited for our new ecosystems. External threats are essentially any attempt by unauthorized users (whether human, robots, extraterrestrials, etc.) to infiltrate and/or take data from a protected system. Internal threats are activities that are attempted by an authorized user – but are not authorized actions for that particular user. An example is a rogue network admin either transferring data to an unauthorized endpoint (piracy) or destruction of data.
The old-fashioned ‘perimeter defense systems’ are no longer appropriate for protection of cloud servers, mobile devices, etc. A particular example of how extensive and interconnected a single ‘session’ can be is given here:
A mobile user opens an app on their phone (say an image editing app) that is ‘free’ to the user. The user actually ‘pays’ for this ‘free’ privilege by donating a small amount of pixels (and time/focus) to some advertising. In the background, the app is providing some basic demographic info of the user, the precise physical location (in many instances), along with other data to an external “ad insertion service”.
This cloud-based service in turn aggregates the ‘avails’ (sorted by location, OS, hardware platform, app type that the user is running, etc.) and often submits these ‘avails’ [with the screen dimensions and animation capabilities] to an online auction system that bids the ‘avails’ against a pool of appropriate ads that are preloaded and ready to be served.
Typically the actual ads are not located on the same server, or even the same country, as either the ad insertion service or the auction service. It’s very common for up to half a dozen countries, clouds and other entities to participate in delivering a single ad to a mobile user.
This highly porous ad insertion system has actually become a recent favorite of hackers and other con games – even without technical breaches it’s an incredibly easy system to game – due to the speed of the transactions and almost impossible ability to monitor in real time many ‘deviations’ are possible… and common.
There are a number of ingenious methods being touted right now to help solve both the actual protection of virtualized and distributed compute environments, as well as to monitor such things as intrusions, breaches and unintended data moves – all things that traditional IT tools don’t address well at all.
I am unaware of a ‘perfect’ solution yet to address either the protection or monitoring aspects, but here are a few ideas: [NOTE: some of these are my ideas, some have been taken up by vendors as a potential product/service. I don’t feel qualified enough to judge the merits of any particular commercial product at this point, nor is the focus of this article on detailed implementations but rather concepts, so I’ll refrain from getting into specific products].
- User endpoint devices (anything from humans’ cellphones to other servers) must be pre-authenticated (using combination of currently well-known identification methods such as MAC address, embedded token, etc.). On top of this basic trust environment, each session is authenticated with a minimum of a two-factor logon scheme (such as biometric plus PIN, certificate plus One Time Token, etc). Once the endpoints are authenticated, a one-time use VPN is established for each connection.
- Endpoint devices and users are combined as ‘profiles’ that are stored as part of a security monitoring application. Each user may have more than one profile: for instance the same user may typically perform (or be allowed to perform by his/her firm’s security protocol) different actions from a cellphone as opposed to a corporate laptop. The actions that each user takes are automatically monitored / restricted. For instance, the VPNs discussed in the point above can be individually tailored to allow only certain kinds of traffic to/from certain endpoints. Actions that fall outside of the pre-established scope, or are outside a heuristic pattern for that user, can either be denied or referred for further authorization.
- Using techniques similar to the SSL methodologies that protect and authenticate online financial transactions, different kinds of certificates can be used to permit certain kinds of ‘transactions’ (with a transaction being either access to certain data, permission to move/copy/delete data, etc.) In a sense it’s a bit like the layered security that exists within the Amazon store: it takes one level of authentication to get in and place an order, yet another level of ‘security’ to actually pay for something (you must have a valid credit card that is authenticated in real time by the clearing houses for Visa/MasterCard, etc.). For instance, a user may log into a network/application instance with a biometric on a pre-registered device (such as fingerprint on an iPhone6 that has been previously registered in the domain as an authenticated device). But if that user then wishes to move several terabytes of a Hollywood movie studio’s content to remote storage site (!!) they would need to submit an additional certificate and PIN.
An Integrated Immune System for Data Security
The goal of a highly efficient and manageable ‘immune system’ for a hyper-connected data infrastructure is for such a system to protect against all possible threats with the least direct supervision possible. Since not only is it impossible for a centralized omniscient monitoring system to handle the incredible number of sessions that take place in even a single modern hyper-network; it’s equally difficult for a single monitoring / intrusion detection device to understand and adapt to the myriad of local contexts and ‘rules’ that define what is ‘normal’ and what is a ‘breach’.
The only practical method to accomplish the implementation of such an ‘immune system’ for large hyper-networks is to distribute the security and protection infrastructure throughout the entire network. Just as in the human body, where ‘security’ begins at the cellular level (with cell walls allowing only certain compounds to pass – depending on the type and location of each cell); each local device or application must have as part of its ‘cellular structure’ a certain amount of security.
As cells become building blocks for larger structures and eventually organs or other systems, the same ‘layering’ model can be applied to IT structures so the bulk of security actions are taken automatically at lower levels, with only issues that deviate substantially from the norm being brought to the attention of higher level and more centralized security detection and action systems.
Another issue of which to be aware: over-reporting. It’s all well and good to log certain events… but who or what is going to review millions of lines of logs if every event that deviates even slightly from some established ‘norm’ is recorded? And even then, that action will only be looking in the rear view mirror.. The human body doesn’t generate any logs at all and yet manages to more or less handle the security for 37.2 trillion cells!
That’s not say that no logs at all should be kept – they can be very useful to help understand breaches and what can be improved in the future – but the logs should be designed with that purpose in mind and recycled as appropriate.
In this very brief overview we’ve discussed some of the challenges and possible solutions to the very different security paradigm that we now have due to the hyper-connected and diverse nature of today’s data ecosystems. As the number of ‘unmanned’ devices, sensors, beacons and other ‘things’ continues to grow exponentially, along with the fact that most of humanity will soon be connected to some degree to the ‘Internet’, the scale of the security issue is truly enormous.
A few ideas and thoughts that can lead to effective, scalable and affordable solutions have been discussed – many of these are new and works in progress but offer at least a partially viable solution as we work forward. The most important thing to take away here is an awareness of how things must change, and to keep asking questions and not assume that the security techniques that worked last year will keep you safe next year.