- Faster, Faster Tech is Running Wild
- Three Ways to Build a New Architecture
Last summer, a group of Google researchers sent Intel a message warning them of an industry-shaking flaw in their chips.
This flaw, which allowed access to ‘partitioned’ kernel memory, allowed the researchers to gain access to passwords and encryption keys, and opened up a critical crack in one of computing’s most basic safeguards.
And the scale of the flaws – dubbed Meltdown and Spectre – has compounded deep fears about security.
How many chips are compromised?
Has anyone been exploiting the flaw?
And given that we can’t correct this problem without new chips – which may be two years away – have we created a new public playground for hackers they never knew existed?
In fact, Meltdown and Spectre are pimples on a cheetah’s bottom.
It’s down to the design priorities that are backed into the system.
At base, you can only have two of the following three to play with in IT design – speed, access and security.
In a world of ever denser, complex and hyper-connected real-time networks of pathologically impatient users, the choice has been foreclosed: go for speed and access, and the devil take the hindmost. It has.
The faster inter-connected networks and their enormous and compounding populations of billions of end point devices operate the more they become accelerants for bad behaviour that can’t be adequately gated and controlled as to what goes in and out.
And guess what? Networks and their components have got to get a lot faster by 2020 if the internet isn’t to get overwhelmed and throttled by the data tsunami and the demands on it to enable conversational computing, self-teaching systems, talk over tap, pervasive mixed reality, autonomous vehicles etc.
As things stand security now needs to become a first concern in system design. Hmm.
At the same time, just as chips exceed transistor counts of 50 billion per die and the spectre of disruptive quantum effects looms, so networks are now so complex that we need to pay heed to the warnings from chaos theorists about complexity and the phenomenon of unexpected and possibly iatrogenic emergent properties.
We live in a world of structural uncertainty where tech is currently running wild.
1) Chip Architectures
Despite breathless talk about carbon nanotubes, spintronics and quantum computing, as far as the mainstream is concerned the big story is the move underway from horizontal to 3D vertical architectures. These involve vertical channels taking the form of pillars of nanowires standing on end.
3D architectures are already the basis of the high performance flash memory chips from Samsung, SK Hynix and Micron. And they may prove to be the basis for high-performance microprocessors as well by the early 2020s, that offer further acceleration, low power draw combined with enhanced firmware security.
A collaboration between IBM, Samsung and Global Foundries <IBM’s original foundry> is scheduled to show its first fruits this year in the form of a processor based on FinFet – field effect transistor architecture – which wraps multiple transistor logic gates round vertical channels.
Google and others will no doubt stress test the firmware security when the first prototypes appear. They won’t have to wait long.
We’ll keep close tabs on this and what Intel is doing, which has hardly been sitting on its hands has in the works.
2) AI-driven security
We also expect to see the emergence of a new cybersecurity industry employing, deep learning, collaborative networks and technology able to detect and frustrate attacks.
Machine learning has been used in the financial sector for some time to identify potential fraud. The large datasets collected on credit card use, for example, have allowed algorithms to learn how to recognise normal behaviour, and in turn to highlight anomalies in the system. IBM researchers working with a large US bank claimed a 15% increase in fraud detection with a 50% reduction in false alarms and a total savings increase of 60%.
The algorithms are still primitive, but once spotted, they are learning to misdirect and confuse, lifting the cost of conducting an attack, sometimes persuading the hacker to move on.
Aside from the big plays – the likes of Symantec and Check Point – the as yet unlisted Palantir is on our watchlist. It’s at the forefront of the business of mining massive, dispersed datasets and connecting the dots of almost invisible but highly significant low frequency data. Valued at $20 billion and funded by PayPal founder Peter Thiel, the CIA’s investment arm Q-tel and ex Soros partner Stanley Druckenmiller.
Darktrace, co-founded by Mike Lynch of Autonomy fame in 2013, has attracted over $100 million in funding from Lynch, KKR and Softbank and is valued at $400 million. Its technology was built in part by former members of M15 and GCHQ and uses unsupervised machine learning algorithms to train themselves to find abnormalities in networks. Darktrace, among others, is leading cybersecurity into a battle between ‘white hat’ and ‘black hat’ AI.
3) Blockchain (or at least more secure databases)
As we said in our recent report, Blockchain is reaching the top of its hype cycle and is poised to enter a trough of disillusionment.
However, a blockchain is only one of many types of distributed ledger.
In the case of bitcoin, the blockchain is shared and validated by a “proof-of-work” system, where miners work through computational tasks, validating each new block on the chain using a cryptographic signature (called a ‘hash’). It’s impossible to change a specific entry secretly, as this alters the hash value and disrupts the data trail.
But there are many, many alternatives to the proof-of-work method of building trust and consensus — depending on the needs and appropriate applications for that industry. We are investigating some of the more promising security applications at the moment.
In truth…these solutions may not be enough.
We also need to see step changes in materials science, edge computing and the development and mass deployment of blockchains that can be tended and secured by AI security.
We’ll probe and monitor what’s gathering on these fronts.
If you would like to keep track of this story or any of our other briefings on chips, ML, or stressed urban masses, then do consider signing up to our free Signum briefings.
Or get in touch if you would like to hear about the four companies at the top of our watchlist.
My email is Michael@signumintel.co.uk.
All information in this briefing is provided in good faith and is accurate to the best of my knowledge. This document is for information purposes only and does not impart financial advice. We recommend you do further due diligence before taking action based on the information in this report.