AI is the new ghost in the network machine
As technology becomes ever more sophisticated, our systems are starting to approach, and in some cases exceed, the complexity of human thought.
The big breakthrough for Artificial Intelligence (AI) was when the IBM Watson cognitive engine was able to win a US gameshow called ‘ Jeopardy’ when pitted against human opponents in 2011, by leveraging natural language understanding and machine learned intelligence for parsing vast data sources in real time and to ‘think’ in a semi-human way. This use of ‘thinking’ technology coupled with massive cloud processing power has evolved to achieve complexity that is now close to, and even beyond, that of the human brain – witness the recent use of supercomputers to calculate the value of Pi to 31 trillion(!) places in Japan.
The idea of human consciousness and intelligence, sitting in a flesh and blood body constrained by the laws of physics, has been described as ‘The Ghost in the Machine’, and quite soon our systems and the networks that support them will have a virtual ‘conscious’ inhabitant too, with AI assisting in overseeing, influencing, and controlling how they function and behave.
The radical de-coupling of the control and management plane from the data plane in Software Defined Networks, has already disaggregate the layer which ‘thinks’ and decides how, where, and when to move the data around the networks, from the underlying ‘doing’ layer which responds, processes the data and moves it as instructed. This allows centralised and cohesive management of the network, fully abstracted from the data and its transit, so a bolt on set of AI tools could be readily positioned to use the same.
OpenFlow protocol to instruct the data plane, and react to threats or execute changes in the same way. This would then supplement the instinct and higher-level contextual awareness of out human network management teams.
As M2M, IoT, and Smart Cities drive ever more traffic with no human involvement in the origin, AI becomes an essential to oversee and manage the traffic behaviours and deal with hostile attacks which could cripple our cities, infrastructure, military, and businesses. Smart AI in networks will have many and varied uses, with some of the major potential adoptions being:
- Predictive attack anticipation with pattern analysis of DOS and DDOS events, threat monitoring and profiling, and proactive mitigation by auto-protection invocation.
- Proactive fixes based on predictive analysis, real-time alarm monitoring, fault history of components and networks with known failure types.
- Real time optimisation of networks to reduce costs by maximising the use of software, licences, virtual machines and containers etc.
- Auto-provisioning of services driven by customer order pipeline, and self-service customer UIs , with parameterised flex up and down capability.
- Insight and intelligent tracking of criminal and terrorist threats.
- Sophisticated traffic and capacity planning based on network loads, trends, historical patterns and potential impact of atypical events.
As AI for augmenting human intelligence and insight becomes all pervasive in our technical ecosystems, we will start to see that the AI ghost is a friendly and helpful spirit in our machines, who will help us make our networks more reliable and efficient in the coming years when benign AI is a part of everyone’s daily experience. If our cities and businesses are safer and work better, then surely that will be hugely beneficial to us all.
Author: Gary Dudbridge
© Gary Dudbridge 2019