Even the general public will be dimly aware of the distinction between the periphery and core of 5G networks from the security perspective, after all the US efforts to expunge Huawei’s from national infrastructures. The periphery, just a popular word for RAN, is considered a lower security risk because it is less susceptible to external attack over the internet and that is why Huawei was sometimes allowed to participate in that but not the core.
More generally, the Huawei saga illustrates the wider dilemma over 5G security, which is that while undoubtedly defences are stronger, so is the exposure to attack. Cybersecurity experts like to talk about the attack surface expanding with 5G, but that is rather a canard. It does not matter if a network or infrastructure has more entry points if at the same time there are new tools that can defend in greater numbers or depth to mitigate that extra exposure.
What matters more is that with 5G, mobile networks finally come of age at the heart of global communications serving mission critical infrastructures such as national electricity grids that come under national security. There are many more incentives for both malign governments and cyber criminals seeking random payments for example to attack 5G networks directly or use it as a conduit to critical systems. As such, 5G may be at greater risk of compromise than its predecessors despite being more secure, because it will draw a lot more fire from serious attackers.
With such considerations in mind, Kevin McNamee, director of threat intelligence at Nokia, recently referred to 5G security as a two-sided coin, with good and bad. On the one hand, 5G is inherently more secure than previous wireless generations, but as McNamee put it vastly increases the attack surface as IoT devices proliferate.
Although we prefer not to refer to attack surfaces, McNamee is certainly right to highlight the IoT as a relatively new and emerging source of risk exposed by mobile networks. It is bringing on a lot of connected devices that are now visible via mobile networks over the internet and that is creating new avenues for old attacks, according to McNamee. He was referring especially to DDoS (distributed denial of service) attacks, which benefit from the scale that can be gained by recruiting large numbers of IoT connected devices in botnets to bombard servers under attack with junk traffic, or pings whose legitimate role is to check whether a remote host is ‘alive’.
A key point here is not that the proliferation of IoT devices changes the game by itself but that it can make new forms of DDoS attack possible. These include carpet bombing, closely related to so called amplification attacks as well as volumetric attacks, designed to overcome the defences or DDoS mitigation techniques that enterprises had employed. With a carpet bombing attack, smaller amounts of traffic are aimed at randomly chosen IP addresses, each one not big enough to trigger DDoS defences.
DDoS defences have mostly relied on detection of large amounts of random traffic aimed at a target server in an attempt to take it temporarily out of service. Carpet bombing attempts to slip beneath the radar by firing at a target from multiple directions at once, whose cumulative affect can be devastating. They may be launched via TCP SYN-ACK reflection attacks, where an attacker spoofs the so called SYN packet, which is normally just a TCP packet sent to another computer requesting a connection for sending subsequent payload data.
The attacking system then replaces its own source IP address with the IP address of the desired ultimate target, before transmitting to an intermediate device identified by the SYN packet’s destination address. Therefore, when the SYN packet arrives at that intermediate destination, instead of being reflected back to the source, it is directed at the target identified by the spoofed IP address. In this way, multiple intermediate destinations each fire packets at the single ultimate target of the DDoS attack, hence the reference to amplification. The degree of amplification depends on how many SYN-ACK retransmits are generated by the reflection service.
The proliferating IoT has led to a spate of such carpet bombing or amplification attacks over the last two years. Such attacks can be damaging in themselves but are also increasingly being used as disruptive smokescreens for more serious penetrative attacks designed to steal data or intellectual property for example.
McNamee attempted to define what needed to be done to shore up 5G against the more concerted attacks from the highly technical proficient players that mobile networks will surely be exposed to, over and above the new measures and tools coming available. He cited monitoring, automation and secure communications as essential areas for attention in securing 5G investments, speaking during a breakout session at the recent virtual (ISC)2 Security Congress 2020. Machine learning techniques can also help defend against carpet bombing attacks by recognizing the associated abnormal traffic profiles developing.
McNamee also cited network slicing as a key element in protecting future 5G networks because of the associated separation between data and applications or services, This can effectively isolate each customer’s slice to resemble a private network. Indeed, operators will be eager to parade the security afforded by network slicing as equivalent to that provided by private networks in order to gain business around the Industrial IoT, for example.
McNamee also highlighted edge computing as a new source of exposure, since placing compute power close to data sources to reduce transmission latency can remove established protections linked to secure data centers. He therefore advocated physical security as well as authentication and robust monitoring for edge compute resources.
At a fundamental level, 5G needs to be equipped not just with the latest security tools but also capability for upgrading protection rapidly to meet arising threats, as well as defense in depth to cater for attacks that inevitably will penetrate outer perimeters. As part of this, the 3GPP SA3 security and privacy group has been developing security standards in tandem with architecture and wireless standards, so that they are integral to the design.
Broadly then 5G inherits the 4G network security architecture but with enhancements designed to meet unexpected attacks. There is more emphasis on stress testing and verifying that new defences work against random attacks. Huawei for example has designed test cases based on threat modelling to verify the effectiveness of new mitigation measures. Huawei calls this the “many eyes and many hands” security verification mechanism.