感知網路跳頻問題研究

Abstract

感知網路(Cognitive radio networks, CRNs)是解決頻譜短缺以及部分 頻普低利用度的關鍵技術。為了保護原始使用者的網路權益，感知網 路內的節點只能相遇(rendezvous)在可用頻道上進行傳送、交換資料、 交換控制封包。其中，可用頻道代表原始使用空出的頻道(idle licensed channel)。然而，實作這樣的相遇技術是艱難的挑戰，因為可用頻道 會隨著使用者的位置、不同時間改變。同時為了減少相遇失敗機率以 及增加傳輸吞吐量，感知網路中傳輸對的通訊訊協定應該設計成可相 遇在所有頻道上(maximizing rendezvous diversity) 以及有快速相遇性質(i.e., minimizing maximum conditional time to rendezvous (MCTTR))。此外，為了充分利用多頻道存取(multi-channel medium access) 的優點，相遇應該平均分攤到不同的頻道上(minimizing channel loading)。在本論文中，我們提出了相遇跳頻演演算法(rendezvous channel hopping algorithms)。所提出的演算法可以運作在時間不同步、收送雙方皆適用的演算法。我們所提出的演算法可提供最好的頻道負擔(minimum channel loading)、相遇不同頻道(maximum rendezvous diversity)、以及目前最短的最大相遇時間(MCTTR)。 然而，現存的跳頻演算法面對阻斷攻擊(jamming attacks) 時卻是 相當無助的，特別是當攻擊者具備感知天線可以監控頻道狀況、 快速跳頻時。許多的抗阻斷攻擊演算法必須是先決定節點共享金 鑰(pre-shared secrets)。在感知網路中，事先決定共享金鑰通常是不 可行的事情，因為鄰近節點可能隨時都會不一樣。因此，抗阻斷 攻擊演算法如何不事先交換共享金鑰成為重要的研究議題。即便如 此，現存的方法，例如：非合作式跳頻演算法(uncoordinated frequency hopping，UFH)，存在著不能同時保證最大相遇時間以及角色共享(也 就是不用事先決定角色是送端或者收端)。特別注意到一點，角色不能 共享很難應用到現實生活中，因為一個角色隨時都有可能會需要收封 包、送封包。也就是有可能會需要扮演收端、送端的角色。在本論文 中，我們提出了一系列的抗阻斷演算法，而這些演算法可以角色共享 以及保證最大的相遇時間。此外，我們分析演算法有高安全度可以抵 抗不同的阻斷攻擊。

Cognitive radio networks (CRNs) have emerged as a critical technique to solve spectrum shortage problem and enhance the utilization of licensed channels. To prevent from interfering with the co-locate incumbent networks, before data transmission, nodes in CRNs should rendezvous on an available channel (i.e., idle licensed channel) for establishing a link or exchanging control information. However, implementing rendezvous is challenging because the availability of channels is time-varying and position-varying. For reducing rendezvous failure and increasing throughput, a node pair in CRN should be able to rendezvous on every licensed channels (i.e., maximizing rendezvous diversity) and rendezvous on an available channel as soon as possible (i.e., minimizing maximum conditional time to rendezvous (MCTTR)). Besides, in order to take full advantage of the frequency diversity of multichannel medium access, rendezvous should be spread out in time and channel (i.e., minimizing channel loading). On the other hand, existing channel hopping algorithms for CRNs are usually vulnerable to jamming attacks, especially when jammers have cognitive radios to perform channel sensing and fast channel switching. Many mitigating approaches for coping with jamming attacks rely on pre-shared secrets (e.g., pre-shared hopping sequences). In CRNs, pre-sharing secrets between senders and receivers is usually impractical (because neighborhood dynamically changes, and receivers of a broadcast may be unknown to the sender). Hence, anti-jamming channel hopping approaches without pre-shared secrets have gained more and more research interests. However, existing approaches, e.g., uncoordinated frequency hopping (UFH), either have unbounded time to rendezvous on an available channel (even no signals of jammers and PUs appear), or require role pre-assignment (SUs should be pre-assigned as a sender or receiver). Role pre-assignment is not applicable to environments where each SU may play a sender and a receiver, simultaneously. In this thesis, we proposed rendezvous channel hopping algorithms, which can be used without time synchronization and role pre-assignment (each node has a pre-assigned role as either a sender or a receiver). Our proposed algorithms have maximum rendezvous diversity and minimum channel loading, and outperform in terms of MCTTR. To enhance the security of communicating pair, we also proposed an algorithm which has high security level to resist various jamming attacks.