AI dominated the news in 2023 and there’s no sign that the hype wave will recede in the next 12 months. While much of the interest was sparked by the existential questions raised by generative AI, CIOs and IT teams continued to struggle with the perennial challenges of rising costs, growing complexity and increasing security risk.
The jury is out on the question of whether AI will result in a long-term net loss or gain for humanity, but its impact on the day-to-day business of IT is already being felt.
Coping with the growing global demands of network traffic
AI will make it easier to generate data – and more data means more network traffic. Even before the AI hype wave broke, global demand for network bandwidth was doubling every two years driven by the entertainment industry, cloud computing and enterprise digital transformation strategies. Companies like Infinera, which makes the optical engines that power the world’s telco networks are finding new ways to respond to rising demand for speed and capacity and the equally powerful demand from network operators for lower cost per bit. The challenge for Infinera and the rest of the optical industry is not just the pace of change, but the immutable laws of physics. We’re nearing the Shannon limit – the maximum rate at which data can be transmitted over a communication channel.
With finite limits on their ability to throw bandwidth at the problem, network operators are looking for more creative solutions – using AI to automate the network for better traffic management, reliability and bandwidth utilisation. Companies like Infinera are increasingly investing in the software that makes networks easier to run.
Optical networking will also have a crucial role in metro areas where 5G and its successors are increasing complexity as well as the need for more terrestrial network capacity.
Expect to see increased demand for pluggable optical engines that meet network operators’ requirements for easily upgradable, standardised solutions without incurring the costs associated with truck rolls and digging up roads.
Solving the challenges of hybrid working
2024 will also be the year that the networking industry gets to grips with the challenge of hybrid working. The work-from-home trend accelerated by the pandemic has changed working practices forever. A majority of staff will work at least part of the time out of the office for the foreseeable future. Since 2020, enterprises have struggled to provide the same experience to staff wherever they work. Unreliable WIFI and home broadband connections create frustration for the user and support overheads for IT. Moving users out of HQ and branch offices also increases security risks as IT loses control over how they connect to the cloud and the data centre.
Centralised network architectures including SD-WAN, VPN, ZTNA and SASE/SSE were not designed for this new world. Deploying them to support the hybrid workforce results in a number of compromises to performance and security. Put simply, IT has to choose whether connections are secure or fast. Solutions that deliver both typically involve additional hardware and increased cost.
Another problem is that users don’t just expect to work from home but from anywhere.
Silicon Valley based start-up Cloudbrink was already working on this problem before the pandemic hit, and came out of stealth mode just over 12 months ago. Its service-based solution moves the networking and security stack out of the office and into the user’s device. Cloudbrink takes advantage of the near unlimited availability edge connectivity in cloud and telco networks to ensure that users are never more than 5ms from the nearest edge. It uses AI to pick the best connection and the best route through the network. It also uses a technique called accelerated pre-emptive packet recovery to overcome the quality issues experienced in large file transfers and real-time video applications such as Zoom and Teams. The result is a 30x improvement in performance over unreliable connections. Importantly, the solution also integrates enterprise-class security with no degradation in performance. Unlike ZTNA providers that rely on dozens, or at most a few hundred, fixed points of presence (PoPs), Cloudbrink uses ephemeral virtual PoPs that can be spun up on demand. That gives it global coverage – with thousands or tens of thousands of PoPs available – but it also creates a moving target for attackers, thanks to the transient nature of the connection.
Advancing Cloud Native networking and security
Another company using Cloud Native technology to solve networking and security problems is NetFoundry. The North Carolina firm is coming at the security challenge from the software end. Traditionally we’ve built software then worked out how to secure it, typically by bolting security onto the networks used by applications. NetFoundry has turned the problem on its head with open-source tools that enable security and networking functions to be incorporated with the application code. Each program becomes, in effect, its own private network enabling internet-exposed firewall ports to be closed, dramatically reducing the attack surface.
The dark networking techniques deployed by companies like NetFoundry and Cloudbrink could help mitigate one of the biggest near-term threats posed by AI: the rise in the volume and ingenuity of cyberattacks.
AI will lower the entry bar for bad actors, making it easier for the unskilled amateur to have a go. The same technology will be used defensively, but the asymmetry in security wars is likely to continue. The fundamental truth will remain that attackers only have to be right once while defenders need to be right 100% of the time.
Major benefits when combining healthy human ingenuity with AI
If we need a reminder that humans are still in the thick of the AI revolution, look no further than Bugcrowd. With headquarters in San Francisco and expanding rapidly internationally, this company bills itself as the only multi-solution crowdsourced cybersecurity platform. It relies on a combination of human expertise and an AI-powered knowledge platform to seek out vulnerabilities in commercial and proprietary software before criminal hackers get there first.
Bugcrowd isn’t just using AI in its work. Increasingly AI is the work – for clients including ChatGPT developer OpenAI. As the company’s founder and chief strategy officer Casey Ellis pointed out, understanding the weaknesses of AI may be critical. “Although AI systems can have well-known vulnerabilities that are found in common web applications, AI technologies like LLMs have introduced unprecedented security challenges that our industry is only beginning to understand and document.” Ellis was speaking last month at Blackhat Europe as Bugcrowd added Large Language Models (LLMs) to its open-source taxonomy for crowdsourcing vulnerabilities.
His colleague Dave Gerry, Bugcrowd’s CEO, provided a timely reminder that the healthy development of AI still depends heavily on its homo sapien creators. “Human ingenuity unleashed by crowdsourced security is the best tool available for meeting AI security goals,” he said.
The hope in 2024 is that the overheated debate about AI settles to become a more nuanced conversation about where the technology can be used to reduce operational costs and risks, raise productivity, create competitive advantage and leveraged in a positive manner to improve people’s lives in and out of the workplace. We’ll know it’s working when we stop talking about it so much.