搜索结果：Preventing

Multi-View Representation Learning (MVRL) aims to learn a unified representation of an object from multi-view data. Deep Canonical Correlation Analysis (DCCA) and its variants share simple formulations and demonstrate state-of-the-art performance. However, with extensive experiments, we observe the issue of model collapse, {\em i.e.}, the performance of DCCA-based methods will drop drastically when training proceeds. The model collapse issue could significantly hinder the wide adoption of DCCA-based methods because it is challenging to decide when to early stop. To this end, we develop NR-DCCA, which is equipped with a novel noise regularization approach to prevent model collapse. Theoretical analysis shows that the Correlation Invariant Property is the key to preventing model collapse, and our noise regularization forces the neural network to possess such a property. A framework to construct synthetic data with different common and complementary information is also developed to compare MVRL methods comprehensively. The developed NR-DCCA outperforms baselines stably and consistently in both synthetic and real-world datasets, and the proposed noise regularization approach can also b

Decoder-only Transformers often struggle with complex reasoning tasks, particularly arithmetic reasoning requiring multiple sequential operations. In this work, we identify representation collapse in the model's intermediate layers as a key factor limiting their reasoning capabilities. To address this, we propose Sequential Variance-Covariance Regularization (Seq-VCR), which enhances the entropy of intermediate representations and prevents collapse. Combined with dummy pause tokens as substitutes for chain-of-thought (CoT) tokens, our method significantly improves performance in arithmetic reasoning problems. In the challenging $5 \times 5$ integer multiplication task, our approach achieves $99.5\%$ exact match accuracy, outperforming models of the same size (which yield $0\%$ accuracy) and GPT-4 with five-shot CoT prompting ($44\%$). We also demonstrate superior results on arithmetic expression and longest increasing subsequence (LIS) datasets. Our findings highlight the importance of preventing intermediate layer representation collapse to enhance the reasoning capabilities of Transformers and show that Seq-VCR offers an effective solution without requiring explicit CoT supervisi

The vulnerability of Deep Neural Networks (DNNs) to adversarial examples has been confirmed. Existing adversarial defenses primarily aim at preventing adversarial examples from attacking DNNs successfully, rather than preventing their generation. If the generation of adversarial examples is unregulated, images within reach are no longer secure and pose a threat to non-robust DNNs. Although gradient obfuscation attempts to address this issue, it has been shown to be circumventable. Therefore, we propose a novel adversarial defense mechanism, which is referred to as immune defense and is the example-based pre-defense. This mechanism applies carefully designed quasi-imperceptible perturbations to the raw images to prevent the generation of adversarial examples for the raw images, and thereby protecting both images and DNNs. These perturbed images are referred to as Immune Examples (IEs). In the white-box immune defense, we provide a gradient-based and an optimization-based approach, respectively. Additionally, the more complex black-box immune defense is taken into consideration. We propose Masked Gradient Sign Descent (MGSD) to reduce approximation error and stabilize the update to i

The paper proposes a Blockchain (BC) system to prevent counterfeiting in health insurance sector. The results show the system strength in terms of achieving data integrity and privacy of data. Moreover, the results show that the consensus algorithm can effectively reduce the total validation time for the proposed system.

The world is currently strongly connected through both the internet at large, but also the very supply chains which provide everything from food to infrastructure and technology. The supply chains are themselves vulnerable to adversarial attacks, both in a digital and physical sense, which can disrupt or at worst destroy them. In this paper, we take a look at two examples of such successful attacks and consider what their consequences may be going forward, and analyse how EU and national law can prevent these attacks or otherwise punish companies which do not try to mitigate them at all possible costs. We find that the current types of national regulation are not technology specific enough, and cannot force or otherwise mandate the correct parties who could play the biggest role in preventing supply chain attacks to do everything in their power to mitigate them. But, current EU law is on the right path, and further vigilance may be what is necessary to consider these large threats, as national law tends to fail at properly regulating companies when it comes to cybersecurity.

Designers incorporate values in the design process that raise risks for vulnerable groups. Persuasion in user interfaces can quickly turn into manipulation and become potentially harmful for those groups in the realm of intellectual disabilities, class, or health, requiring proactive responsibility approaches in design. Here we introduce the Capability Sensitive Design Approach and explain how it can be used proactively to inform designers' decisions when it comes to evaluating justice in their designs preventing the risk of manipulation.

Many conferences rely on paper bidding as a key component of their reviewer assignment procedure. These bids are then taken into account when assigning reviewers to help ensure that each reviewer is assigned to suitable papers. However, despite the benefits of using bids, reliance on paper bidding can allow malicious reviewers to manipulate the paper assignment for unethical purposes (e.g., getting assigned to a friend's paper). Several different approaches to preventing this manipulation have been proposed and deployed. In this paper, we enumerate certain desirable properties that algorithms for addressing bid manipulation should satisfy. We then offer a high-level analysis of various approaches along with directions for future investigation.

The reinforcement learning (RL) problem is rife with sources of non-stationarity, making it a notoriously difficult problem domain for the application of neural networks. We identify a mechanism by which non-stationary prediction targets can prevent learning progress in deep RL agents: \textit{capacity loss}, whereby networks trained on a sequence of target values lose their ability to quickly update their predictions over time. We demonstrate that capacity loss occurs in a range of RL agents and environments, and is particularly damaging to performance in sparse-reward tasks. We then present a simple regularizer, Initial Feature Regularization (InFeR), that mitigates this phenomenon by regressing a subspace of features towards its value at initialization, leading to significant performance improvements in sparse-reward environments such as Montezuma's Revenge. We conclude that preventing capacity loss is crucial to enable agents to maximally benefit from the learning signals they obtain throughout the entire training trajectory.

Context] Problems in Requirements Engineering (RE) can lead to serious consequences during the software development lifecycle. [Goal] The goal of this paper is to propose empirically-based guidelines that can be used by different types of organisations according to their size (small, medium or large) and process model (agile or plan-driven) to help them in preventing such problems. [Method] We analysed data from a survey on RE problems answered by 228 organisations in 10 different countries. [Results] We identified the most critical RE problems, their causes and mitigation actions, organizing this information by clusters of size and process model. Finally, we analysed the causes and mitigation actions of the critical problems of each cluster to get further insights into how to prevent them. [Conclusions] Based on our results, we suggest preliminary guidelines for preventing critical RE problems in response to context characteristics of the companies.

Context. Technical Debt (TD) is a metaphor for technical problems that are not visible to users and customers but hinder developers in their work, making future changes more difficult. TD is often incurred due to tight project deadlines and can make future changes more costly or impossible. Project Management usually focuses on customer benefits and pays less attention to their IT systems' internal quality. TD prevention should be preferred over TD repayment because subsequent refactoring and re-engineering are expensive. Objective. This paper evaluates a framework focusing on both TD prevention and TD repayment in the context of agile-managed projects. The framework was developed and applied in an IT unit of a publishing house. The unique contribution of this framework is the integration of TD management into project management. Method. The evaluation was performed as a comparative case study based on ticket statistics and two structured surveys. The surveys were conducted in the observed IT unit using the framework and a comparison unit not using the framework. The first survey targeted team members, the second one IT managers. Results. The evaluation shows that in this IT unit,

Collective decoherence is possible if the departure between quantum bits is smaller than the effective wave length of the noise field. Collectivity in the decoherence helps us to devise more efficient quantum codes. We present a class of optimal quantum codes for preventing collective amplitude damping to a reservoir at zero temperature. It is shown that two qubits are enough to protect one bit quantum information, and approximately $L+ 1/2 \log_2((πL)/2)$ qubits are enough to protect $L$ qubit information when $L$ is large. For preventing collective amplitude damping, these codes are much more efficient than the previously-discovered quantum error correcting or avoiding codes.

SQL injection attacks, a class of injection flaw in which specially crafted input strings leads to illegal queries to databases, are one of the topmost threats to web applications. A Number of research prototypes and commercial products that maintain the queries structure in web applications have been developed. But these techniques either fail to address the full scope of the problem or have limitations. Based on our observation that the injected string in a SQL injection attack is interpreted differently on different databases.Injection attack is a method that can inject any kind of malicious string or anomaly string on the original string. Pattern matching is a technique that can be used to identify or detect any anomaly packet from a sequential action. Most of the pattern based techniques are used static analysis and patterns are generated from the attacked statements. In this paper, we proposed a detection and prevention technique for preventing SQL Injection Attack using AhoCorasick pattern matching algorithm. In this paper, we proposed an overview of the architecture. In the initial stage evaluation, we consider some sample of standard attack patterns and it shows that the p

In recent years, the interest in interpretable classification models has grown. One of the proposed ways to improve the interpretability of a rule-based classification model is to use sets (unordered collections) of rules, instead of lists (ordered collections) of rules. One of the problems associated with sets is that multiple rules may cover a single instance, but predict different classes for it, thus requiring a conflict resolution strategy. In this work, we propose two algorithms capable of finding feature-space regions inside which any created rule would be consistent with the already existing rules, preventing inconsistencies from arising. Our algorithms do not generate classification models, but are instead meant to enhance algorithms that do so, such as Learning Classifier Systems. Both algorithms are described and analyzed exclusively from a theoretical perspective, since we have not modified a model-generating algorithm to incorporate our proposed solutions yet. This work presents the novelty of using conflict avoidance strategies instead of conflict resolution strategies.

Nowadays, denial of service (DoS) attacks represent a significant fraction of all attacks that take place in the Internet and their intensity is always growing. The main DoS attack methods consist of flooding their victims with bogus packets, queries or replies, so as to prevent them from fulfilling their roles. Preventing DoS attacks at network level would be simpler if end-to-end strong authentication in any packet exchange was mandatory. However, it is also likely that its mandatory adoption would introduce more harm than benefits. In this paper we present an end-point addressing scheme and a set of security procedures which satisfy most of network level DoS prevention requirements. Instead of being known by public stable IP addresses, hosts use ephemeral IP Identifiers cryptographically generated and bound to its usage context. Self-signed certificates and challenge-based protocols allow, without the need of any third parties, the implementation of defenses against DoS attacks. Communication in the open Internet while using these special IP addresses is supported by the so-called Map/Encap approaches, which in our point of view will be sooner or later required for the future In