Search for: All records

Spectrum Access Systems (SASs) and similar systems coordinate access to shared radio frequency bands to efficiently allocate the use of spectrum between users in a locality. To fill the need for dense spectrum occupancy information, SASs will utilize crowdsourced data from nodes outside the SAS’s control. This crowdsourcing of data, however, makes the SAS vulnerable to many types of attacks. The attacks covered in this paper include copying and manipulating existing data to create a Spectrum Sensing Data Falsification (SSDF) attack. We propose methods to identify two categories of easily implemented SSDF attacks and show the proposed methods to be both effective and efficient. Further, we recommend that the proposed techniques be used in conjunction with other SSDF thwarting methods that use statistics, probability, or machine learning, and can identify a wider range of SSDF attacks, albeit more slowly and less reliably than the proposed methods can identify the specific types of SSDF attacks for which they are effective. Our findings demonstrate the feasibility of discerning diverse forms of manipulated data while maintaining pace with the influx of incoming data. The ability to identify manipulated data rapidly without imposing undue strain on a centrally aggregated system helps reduce the number of ways to create a potentially successful SSDF attack and increases the accuracy of determining the radio transmission activity of a Primary User (PU). Several methods are explored and evaluated for identifying copied or manipulated spectrum data. We recommend utilizing an exact match identification algorithm with Elasticsearch to search for exact copies of spectrum data. Additionally, we recommend utilizing a cosine similarity function with Elasticsearch to search for manipulated spectrum data and exact copies when sufficient computational resources are available.