Given that distributed systems face adversarial behaviors such as eavesdropping and cyberattacks, how to ensure the evidence fusion result is credible becomes a must-be-addressed topic. Different from traditional research that assumes nodes are cooperative, we focus on three requirements for evidence fusion, i.e., preserving evidence's privacy, identifying attackers and excluding their evidence, and dissipating high-conflicting among evidence caused by random noise and interference. To this end, this paper proposes an algorithm for credible evidence fusion against cyberattacks. Firstly, the fusion strategy is constructed based on conditionalized credibility to avoid counterintuitive fusion results caused by high-conflicting. Under this strategy, distributed evidence fusion is transformed into the average consensus problem for the weighted average value by conditional credibility of multi-source evidence (WAVCCME), which implies a more concise consensus process and lower computational complexity than existing algorithms. Secondly, a state decomposition and reconstruction strategy with weight encryption is designed, and its effectiveness for privacy-preserving under directed graphs is guaranteed: decomposing states into different random sub-states for different neighbors to defend against internal eavesdroppers, and encrypting the sub-states' weight in the reconstruction to guard against out-of-system eavesdroppers. Finally, the identities and types of attackers are identified by inter-neighbor broadcasting and comparison of nodes' states, and the proposed update rule with state corrections is used to achieve the consensus of the WAVCCME. The states of normal nodes are shown to converge to their WAVCCME, while the attacker's evidence is excluded from the fusion, as verified by the simulation on a distributed unmanned reconnaissance swarm.

翻译：鉴于分布式系统面临窃听和网络攻击等对抗行为，如何确保证据融合结果的可信性成为一个亟待解决的课题。不同于传统研究假设节点具有合作性，本文聚焦于证据融合的三个核心要求：保护证据隐私、识别攻击者并排除其证据、以及消除由随机噪声和干扰引起的高冲突证据。为此，本文提出一种面向网络攻击的可信证据融合算法。首先，基于条件化可信度构建融合策略，以避免高冲突导致的违反直觉的融合结果。在此策略下，分布式证据融合转化为多源证据条件可信度加权平均值（WAVCCME）的平均一致性问题，相较于现有算法，该过程具有更简洁的共识流程和更低计算复杂度。其次，设计了一种带权重加密的状态分解与重构策略，并证明了其在有向图下隐私保护的有效性：通过将状态分解为面向不同邻居的随机子状态以防御内部窃听者，并在重构过程中加密子状态权重以抵御系统外窃听者。最后，通过邻居间广播和节点状态比较识别攻击者身份与类型，并采用所提出的带状态修正的更新规则实现WAVCCME的共识。仿真实验以分布式无人侦察集群为验证场景，结果表明正常节点状态收敛于其WAVCCME，而攻击者证据被有效排除在融合过程之外。