Spectrum Sensing Data Falsification (SSDF)

A malicious node persistently reports a local decision of '1' (occupied) regardless of the actual spectrum state. This strategy aims to maximize the global probability of false alarm, convincing the fusion center that the channel is perpetually busy. The result is a denial-of-service against the secondary network, forcing legitimate cognitive radios to vacate usable spectrum. This attack is trivially easy to execute but highly effective against simple majority-rule fusion schemes.

The adversary consistently reports a local decision of '0' (vacant) irrespective of ground truth. This strategy is designed to blind the fusion center to the presence of a primary user, maximizing the global probability of misdetection. The consequence is catastrophic: secondary users are permitted to transmit, causing harmful interference to the licensed primary user. This attack directly violates the cardinal rule of cognitive radio and exposes the network operator to regulatory penalty.

The attacker flips its local sensing report with a fixed probability P_mal, independently of the true spectrum state. This introduces stochastic noise into the fusion process, degrading the global Receiver Operating Characteristic (ROC) curve. Unlike deterministic attacks, random falsification is harder to attribute to a specific node using simple consistency checks. The attacker trades off immediate catastrophic impact for stealth and persistence, slowly eroding the network's sensing reliability over time.

A more sophisticated strategy where the attacker accurately senses the spectrum but intentionally inverts its report before transmission. If the primary user is present, it reports '0'; if absent, it reports '1'. This is a coherent adversarial strategy that is maximally disruptive for a single node. Because the attacker's sensing capability is intact, it can adapt its falsification to the dynamic environment, making it particularly damaging in networks with a small number of cooperating nodes.

Multiple compromised nodes coordinate their falsification to create a consistent, mutually reinforcing false narrative. For example, a group of M attackers might all report '1' simultaneously to overpower a K-out-of-N fusion rule. This coordinated action defeats simple outlier detection, as the malicious reports form a statistical cluster. The attack leverages the cooperative nature of the network against itself, representing the most dangerous threat model requiring advanced reputation management to counter.

An intelligent, stealthy strategy where the attacker alternates between periods of honest reporting and periods of falsification. The node builds a high reputation score during 'on' (honest) phases, then exploits that trust during 'off' (malicious) phases to inflict targeted damage at critical moments. This attack is specifically designed to evade dynamic reputation management systems that rely on long-term historical consistency, as the attacker's average behavior may appear statistically benign.

The formal name for an SSDF attack, derived from the Byzantine Generals Problem in distributed computing. A malicious node sends arbitrary or conflicting information to different parts of the network to prevent consensus. In CSS, this means a node reports falsified sensing data—sending '1' to some fusion rules and '0' to others, or strategically flipping bits—to corrupt the global spectrum occupancy decision.

A distinct but related physical-layer attack where a malicious actor transmits a signal mimicking a licensed primary user. Unlike SSDF, which corrupts the sensing data reporting channel, PUE attacks the sensing channel itself by creating a false signal. Legitimate secondary users detect the fake primary user and erroneously vacate the spectrum, creating a denial-of-service condition.

The central processing node that aggregates local sensing reports to make a global spectrum occupancy decision. It is the primary target of SSDF attacks. Fusion centers implement rules like:

• Hard Decision Fusion: Binary '1' or '0' votes
• Soft Decision Fusion: Raw energy level statistics
• K-out-of-N Rule: Requires K positive votes from N nodes The fusion rule's robustness directly determines SSDF resilience.

Two fusion strategies with different SSDF vulnerabilities:

• Hard Decision Fusion: Nodes send binary decisions. A single flipped bit from a malicious node can directly alter the global vote. Simpler but more vulnerable.

• Soft Decision Fusion: Nodes send quantized test statistics. An attacker must inject statistically plausible false values, making attacks harder to execute but still possible with knowledge of the fusion algorithm.

A decentralized alternative to fusion center architectures. Nodes iteratively exchange information only with neighbors and run a distributed consensus algorithm to converge on a common decision. This eliminates the single point of failure that SSDF attacks exploit, but introduces new attack vectors like influencing the consensus gradient through carefully crafted false updates to neighboring nodes.