To examine cognitive radio from a security perspective we again return to its definition. One of the strongest features of a cognitive radio is the fact it takes in and processes so much information. In other words it observes much. It then adapts its behaviour on the basis of the observations through making appropriate decisions and taking action. The observations include an understanding of the environment in which the radio operates, an understanding of the communication requirements of the user(s), an understanding of the regulatory policies which apply to it and an understanding of its own capabilities. we shouldn’t forget that cognitive radio is ‘all about understanding context’.
While other radios use information to trigger adaptive behavior the breadth of information that is used in a cognitive radio is more vast. Some information is particularly crucial. For example, information about the regulatory policies that apply is vital in ensuring a radio behaves correctly. Information about the radio environment is essential if interference is to be properly managed, while information about user requirements might be of lesser or greater importance. Whatever the importance though, all action springs from the information the radio gets. Hence if the observations are wrong then pretty much everything else will be wrong too! In other words, the very strength of a cognitive radio, i.e. its ability to be aware of its surroundings and circumstances through the gathering of information and observations in order to decide how to act, is also a huge weakness. The dependency of a cognitive radio on observations opens up a plethora of opportunities for security attacks. Very simply put, if false data can be generated or if data can be tampered with, the behavior of the cognitive radio can be controlled.
Physical fakes
Attackers can generate signals that are correctly observed by cognitive radios but are in fact only generated to ‘confuse’ the cognitive radio. This type of attack makes sense especially in the context of dynamic spectrum access networks, in which the cognitive radio senses the presence of an incumbent or primary user. The malicious terminal emits signals that emulate the characteristics of the incumbent. This malicious terminal can cause the cognitive radio to think that there is no unoccupied spectrum. This will either result in the cognitive radio not transmitting at all or moving unnecessarily to another frequency. Whether this is a likely event or not remains to be seen. The attack takes some effort and equipment to stage. And in one sense, the more agile the radio, the more resistant it is to such an attack as it will be able to use a plethora of other frequencies (unless in the unlikely event the attacker occupies all frequencies). Should this kind of attack turn out to be a strong possibility, it is possible to create systems that are robust to this. Cognitive radios may have to ensure that any feature-based sensing uses incumbent/primary characteristics that are non-forgeable in detecting their presence.
Physical degrading of performance to stop sensing
Attackers can raise the noise floor in the vicinity of a cognitive radio to make it impossible to detect signals that should normally be detected. On the one hand this seems like a problem for any radio system, i.e. it can be jammed. However, the issue here is that it is just the ability of the cognitive radio to sense that needs to be disrupted. This will take much less effort than the jamming of its ability to communicate. The problem with these jamming attacks is that it may be very difficult to distinguish between malicious intent and a badly behaving cognitive neighbour. For example, if a neighbouring cognitive node ends up unintentionally clipping the signal it is transmitting, and as a result causes alarming amounts of spectral regrowth around its transmission profile, it can cause problems to other cognitive radios. Spectrum policing/monitoring may be needed to ensure that there are no malicious or misbehaving nodes. And, a point we will return to later, if it is possible for the cognitive radio to rely on multiple sources of information from other parties, there may be a means of getting the information it needs when it cannot generate it itself.
Physical tampering with data
Attackers can physically tamper with stored data such as policy databases. Any tampering with the regulatory policies that are used by the radio is of grave concern. It is, however, perhaps safe to assume that any policy database would be well secured. There also may be ways of ensuring content integrity. Content integrity means that the receiving party can be sure that the transferred information is exactly what the other party originally sent. Integrity can be established by computing a message checksum or digest and attaching it in every message before the message is sent. The receiving party can then re-compute the digest and compare it with the original digest. The cognitive radios can also authenticate policies before enforcing them, to ensure they came from the correct party. Reiterating the point that was made at the outset of this section, in radios in which data plays such a great role in understanding context and in deciding actions, all data needs to be treated with care and to be secure.
