The widespread adoption of TLS 1.3 and QUIC renders payload content invisible,shifting traffic analysis toward side-channel features.However,rigorous justification for"why side-channel leakage is inevitable in encrypted com⁃munications"has long been lacking.This paper establishes a strict foundation from information theory and system design by constructing a formal model Σ=(Γ,Ω),where the encrypted communication system Γ=(A,Π,Φ,N)describes the causal chain of"application generation-protocol encapsulation-encryption transformation-network transmission",and the observa⁃tion model Ω characterizes external observation capabilities.This framework abstracts the complete communication process as a causally measurable Markov chain X→ΞA →ΞP →ΞC→ΞN→Y,enabling the mutual information between semantic variables and observable features to be rigorously defined.Based on the composite channel structure,data processing in⁃equality,and stable propagation of bounded Lipschitz statistics,we propose and prove the"Side-Channel Existence Theo⁃rem":for distinguishable semantic pairs,under the conditions that the system satisfies mapping non-degeneracy(bounded metric expectation E[d(zP,zN)∣X]≤C),protocol-layer statistical distinguishability(expectation difference≥(△ˉ)),Lipschitz continuity of statistics,observation non-degeneracy(preservation ratio ρ>0),and the distinguishability propagation condi⁃tion(C<(△ˉ)/2Lφ),the mutual information I(X;Y)between observed features and semantic variables is necessarily strictly pos⁃itive with an explicit lower bound.The corollary demonstrates that in efficiency-prioritized multi-semantic systems,side-channel leakage is inevitable as long as at least one pair of applications is statistically distinguishable.Three key factors jointly determine the leakage boundary:the mapping non-degeneracy constant C is constrained by efficiency require⁃ments,reflecting practical demands such as bandwidth and latency;semantic distinguishability(△ˉ)stems from application diversity,embodying inherent differences in statistical characteristics across applications;and observation non-degeneracy ρ is determined by analyst capabilities.This paper further establishes a quantitative connection from information-theoretic lower bounds to classification accuracy through the chain of total variation and Chernoff information bounds,revealing the inevitability that multiple observations cause recognition error rates to decay exponentially.Theoretical analysis shows that reducing leakage faces a trilemma:increasing metric deviation requires sacrificing efficiency,reducing semantic dis⁃tinguishability disrupts application functionality,while observation non-degeneracy is controlled by analysts.Therefore,side channels are not incidental flaws in protocol implementations but inherent properties of network communication sys⁃tems subject to practicality constraints,and the correct engineering objective is a constrained optimization problem that minimizes leakage under given efficiency constraints.This paper establishes,for the first time,a rigorous information-the⁃oretic foundation for encrypted traffic side channels,providing verifiable predictions for attack feasibility,quantifiable per⁃formance benchmarks for defense mechanisms,and mathematical basis for engineering decisions on efficiency-privacy tradeoffs.
Nowadays many algorithms have appeared to solve many-objective optimization problems (MaOPs), yet the balance between convergence and diversity is still an open issue. In this paper, we propose a dynamic decomposition and hyper-distance based many-objective evolutionary algorithm named DHEA. On one hand, to maximize the diversity of the population, we use dynamic decomposition to decompose the whole population into multiple clusters. Specifically, first find pivot solutions according to the distribution of the population through the max–min-angle strategy, and then, assign solutions into different clusters according to their distances to pivot solutions. On the other hand, to select solutions from each cluster with balanced convergence and diversity, we propose hyper-distance based angle penalized distance for fitness assignment. Specifically, first compute the distance of solutions to the hyperplane and to the pivot solution to measure convergence and diversity, respectively, and then select the solution with the smallest fitness value. Hyper-distance, as convergence-related component, alleviates the bias towards problems with concave PFs. Besides, to promote convergence, the concept of knee points is introduced to mating selection. Through comparison with nine algorithms on 27 test problems, DHEA is validated to be effective and competitive to deal with MaOPs with different types of Pareto fronts and stable on different numbers of objectives.
In the scenario of imperfect channel state information(CSI)and imperfect successive interference cancellation(SIC),the problem of robust secure resource allocation in reconfigurable intelligent surface(RIS)assisted multi-user non-orthogonal multiple access(NOMA)is studied.Considering energy users as potential eavesdropping nodes in the network,Under constraints on the quality of service(QoS)and maximum eavesdropping rate,a transmit power minimization problem is formulated for the network where there are two types of users(information user and energy user).First,SIC decoding order is determined based on the channel gain of information users,then the S-procedure is used to transform the non-convex constraints with channel uncertainty,and then the optimization problem is decomposed into two sub-problems.Finally,the alternate optimization method based on semi-definite relaxation is used to iteratively solve the above sub-problems.The simulation results show that in the case of imperfect channel state information,optimizing the decoding sequence of successive interference cancellation can effectively reduce the transmit power of the base station.
Graph anomaly detection is essential for identifying irregular patterns and outliers within complex network structures in domains like social networks, cybersecurity, finance, and transportation systems. It helps detect security breaches, fraud, and errors, improving decision-making and system reliability. Although current methodologies have advanced the unsupervised detection of graph anomalies, they frequently fail to fully address the nuanced specificities of graph anomalies, such as anomalous nodes, edges, and subgraphs. To overcome this limitation, the study presents MulDualGNN, a unique dual-channel heterogeneous graph anomaly detection framework. MulDualGNN incorporates a global substructure-aware GNN and a local substructure-aware GNN to capture both global and local substructure properties for accurate anomaly detection. Our model incorporates a multi-hypersphere learning target function, which includes macroscopic and mesoscopic hyperspheres. These hyperspheres measure abnormal nodes that deviate from most normal nodes in the entire graph and community structure, respectively. To overcome the model collapse problem in multi-hypersphere learning, our model utilizes the EmbSim similarity function to optimize the training target. The effectiveness and performance advantages of the proposed method are evaluated through extensive experiments on five datasets. The results demonstrate the superior performance of our approach in graph anomaly detection tasks.
In response to escalating cybersecurity threats, this study aims to develop a lightweight deep learning model for efficient and accurate detection and classification of malicious code. To achieve this goal, the study introduces TriCh-RepNet, a novel network architecture that balances performance and resource utilization. The methodology involves innovatively transforming malicious code representations into image channels through a three-channel mapping technique, thereby enhancing information richness and discriminatory power. By integrating the strengths of Convolutional Neural Networks (CNNs) and Transformers, TriCh-RepNet creates a streamlined framework that optimizes network connections, reduces memory access overhead, and boosts overall efficiency. Additionally, the combination of linear training time over-parameterization and large kernel convolution techniques minimizes the model's parameter count and computational load. The model achieves accuracy rates of 99.47% and 97.51% on the Kaggle and DataCon datasets, respectively. Compared to existing methodologies, this approach stands out in terms of performance, resource consumption, and versatility. It offers a viable and efficient solution, especially in resource-constrained or real-time environments. A novel and effective solution for malicious code detection and categorization.
