Attack Step Research Articles

Adaptive Cross-Modal Transferable Adversarial Attacks From Images to Videos.

The cross-model transferability of adversarial examples makes black-box attacks to be practical. However, it typically requires access to the input of the same modality as black-box models to attain reliable transferability. Unfortunately, the collection of datasets may be difficult in security-critical scenarios. Hence, developing cross-modal attacks for fooling models with different modalities of inputs would highly threaten real-world DNNs applications. The above considerations motivate us to investigate cross-modal transferability of adversarial examples. In particular, we aim to generate video adversarial examples from white-box image models to attack video CNN and ViT models. We introduce the Image To Video (I2V) attack based on the observation that image and video models share similar low-level features. For each video frame, I2V optimizes perturbations by reducing the similarity of intermediate features between benign and adversarial frames on image models. Then I2V combines adversarial frames together to generate video adversarial examples. I2V can be easily extended to simultaneously perturb multi-layer features extracted from an ensemble of image models. To efficiently integrate various features, we introduce an adaptive approach to re-weight the contributions of each layer based on its cosine similarity values of the previous attack step. Experimental results demonstrate the effectiveness of the proposed method.

Stronger and Transferable Node Injection Attacks

Despite the increasing popularity of graph neural networks (GNNs), the security risks associated with their deployment have not been well explored. Existing works follow the standard adversarial attacks to maximize cross-entropy loss within an L-infinity norm bound. We analyze the robustness of GNNs against node injection attacks (NIAs) in black-box settings by allowing new nodes to be injected and attacked. In this work, we propose to design stronger and transferable NIAs. First, we propose margin aware attack (MAA) that uses a maximum margin loss to generate NIAs. We then propose a novel margin and direction aware attack (MDA) that diversifies the initial directions of MAA attack by minimizing the cosine similarity of the injected nodes with respect to their respective random initialization in addition to the maximization of max-margin loss. This makes the NIAs stronger. We further observe that using L2 norm of gradients in the attack step leads to an enhanced diversity amongst the node features, thereby further enhancing the strength of the attack. We incorporate transferability in NIAs by perturbing the surrogate model before generating the attack. An analysis of eigen spectrum density of the hessian of the loss emphasizes that perturbing the weights of the surrogate model improves the transferability. Our experimental results demonstrate that the proposed resilient node injection attack (R-NIA) consistently outperform PGD by margins about 7-15% on both large and small graph datasets. R-NIA is significantly stronger and transferable than existing NIAs on graph robustness benchmarks.

N-heterocyclic carbene catalyzed [2 + 3] annulation reaction for the synthesis of trifluoroethyl 3,2'-spirooxindole γ-lactam.

Asymmetric catalytic processes promoted by N-heterocyclic carbenes (NHCs) hold great potential for the sustainable preparation of chiral molecules. However, catalyzing the reactions by manipulating the reactive intermediates is challenging. We report herein that the known NHC-catalyzed [3 + 2] annulation reaction between ketimine and enal can also be turned into a [2 + 3] annulation reaction for the highly enantioselective direct synthesis of trifluoroethyl 3,2'-spirooxindole γ-lactams (4) through timely catalysis of the intermediates. DFT calculations revealed that this transformation included the key step of the nucleophilic attack of the Breslow intermediate M2 derived from NHC and enal (2) to the unattacked ketimine (1). Our study demonstrates that it is possible to tune the desired selectivities through the dynamic catalysts of the reactive intermediates.

Mononuclear Non-Heme Manganese-Catalyzed Enantioselective cis-Dihydroxylation of Alkenes Modeling Rieske Dioxygenases.

The practical catalytic enantioselective cis-dihydroxylation of olefins that utilize earth-abundant first-row transition metal catalysts under environmentally friendly conditions is an important yet challenging task. Inspired by the cis-dihydroxylation reactions catalyzed by Rieske dioxygenases and non-heme iron models, we report the biologically inspired cis-dihydroxylation catalysis that employs an inexpensive and readily available mononuclear non-heme manganese complex bearing a tetradentate nitrogen-donor ligand and aqueous hydrogen peroxide (H2O2) and potassium peroxymonosulfate (KHSO5) as terminal oxidants. A wide range of olefins are efficiently oxidized to enantioenriched cis-diols in practically useful yields with excellent cis-dihydroxylation selectivity and enantioselectivity (up to 99% ee). Mechanistic studies, such as isotopically 18O-labeled water experiments, and density functional theory (DFT) calculations support that a manganese(V)-oxo-hydroxo (HO-MnV═O) species, which is formed via the water-assisted heterolytic O-O bond cleavage of putative manganese(III)-hydroperoxide and manganese(III)-peroxysulfate precursors, is the active oxidant that effects the cis-dihydroxylation of olefins; this is reminiscent of the frequently postulated iron(V)-oxo-hydroxo (HO-FeV═O) species in the catalytic arene and alkene cis-dihydroxylation reactions by Rieske dioxygenases and synthetic non-heme iron models. Further, DFT calculations for the mechanism of the HO-MnV═O-mediated enantioselective cis-dihydroxylation of olefins reveal that the first oxo attack step controls the enantioselectivity, which exhibits a high preference for cis-dihydroxylation over epoxidation. In this study, we are able to replicate both the catalytic function and the key chemical principles of Rieske dioxygenases in mononuclear non-heme manganese-catalyzed enantioselective cis-dihydroxylation of olefins.

Resilient control strategy and attack detection on platooning of smart vehicles under DoS attack

Intent on the smart vehicles of a platoon to follow their correct routes, they must exchange their state information with each other based on a communication graph. Meanwhile, malicious factors like Denial-of-Service (DoS) attacks, one of the major types of cyber-attacks, can affect vehicles and divert them from their correct routes. Also, spreading the erroneous data induced by a DoS attack from the attacked agent gradually destabilizes the platoon. Therefore, preserving all vehicles’ safety is a vital issue for the platoon. In this paper, in the first step, detection and measurement of the DoS attack modeled by a time-varying delay are conducted by exploiting two incremental counters through a DoS detection and measurement algorithm. Also, supplementary to this, a novel vehicular resilient control strategy based on switching systems is proposed to retrieve the attacked agent to the leader–follower consensus. In the end, the capability of the proposed algorithms will be indicated by presenting an illustrative case study.

Threat-Based Simulation of Data Exfiltration Toward Mitigating Multiple Ransomware Extortions

Network-based attacks and their mitigation are of increasing importance in our ever-connected world. Often network-based attacks address valuable data, which the attacker either encrypts to extort ransom or steals to make money reselling, or both. After the infamous WannaCry and NotPetya ransomware attacks in 2017, companies stepped up their cyber defenses. More emphasis was placed on backup and recovery processes so that even when files were destroyed, organizations had copies for quick recovery. However, cyber criminals have also adapted their methods. Instead of simply encrypting files, double extortion ransomware now exfiltrates the data first, before encrypting it. As a consequence, the early detection and prevention of data exfiltration is one of today’s major challenges of institutions connected to the Internet. If attempts to illegal data exfiltration are successfully detected, the attacked institution should address a probable subsequent encryption attack step as well. In particular, valuable business assets must be checked for unauthorized access and need to be protected. However, due to the bulk of network traffic and persistent data, automation is a key requirement to successfully defend contemporary threats. The main goal of this article is to present a concept and its initial evaluation to achieve automation of data exfiltration mitigation in a targeted manner. Our concept consists of two main steps. Based on recognized international approaches used in cyber threat intelligence, an automatic procedure on the base of the MITRE Adversarial Tactics, Techniques Common Knowledge (ATT&CK) framework for deriving current threats with respect to data exfiltration is presented in the first place. In the spirit of the Digital Threats: Research and Practice (DTRAP) forum, a practical approach is chosen in addition to the theory in this manner. Our evaluation reveals that we are able to automatically identify the most relevant recent risks of unauthorized data exfiltration. In our second step, we present the design of a simulation gear based on the attacks extracted from the MITRE ATT&CK framework. The aim is to simulate the greatest threats before they actually occur in the operational environment. The strict focus on the threats of data exfiltration characterizes our solution and makes our approach an ideal addition to existing solutions. We provide an evaluation of this initial simulation concept and its underlying technology for the implementation to show that we are on the right track.

Optimization and theoretical analysis of lipase-catalyzed enzymatic esterification of glycerol for efficient glycerides synthesis

Eversa® Transform 2.0 is a lipase derived from Thermomyces lanuginosus produced by a genetically modified microorganism. It has a low production cost, high substrate specificity, and high catalytic activity in organic synthesis. The Taguchi method was used to evaluate the best conditions for the enzymatic esterification of glycerol with acetic acid. A yield of 84.8% was obtained, under the optimal conditions (temperature = 40 °C; molar ratio glycerol/acid = 1:1; biocatalyst = 15% w/w; time = 12.5 h). After the statistical analysis, the temperature was found to be the most significant parameter influencing the reaction conversion. A theoretical study was carried out to generate a homology model of the enzyme, based on other natural lipases. Molecular docking, molecular dynamics, and QM/MM simulations were applied to understand the mechanism of esterification and to derive thermodynamic and kinetic data. The nucleophilic attack step was identified as the rate-limiting step for both acylation (13.1 kcal/mol) and deacylation (13.8 and 12.9 kcal/mol) reaction mechanisms. Although the enzyme is capable of esterifying all three alcohol groups of glycerol, the esterification of the primary alcohols is thermodynamically more favorable (5 kcal/mol), especially at higher temperatures, than their secondary counterpart.

Metal-free recycling of waste polyethylene terephthalate mediated by TBD protic ionic salts: the crucial role of anionic ligands.

The structure-activity relationships of 1,5,7-triazabicyclo[4.4.0]dec-5-ene (TBD) based protic ionic salts for polyethylene terephthalate (PET) glycolysis by ethylene glycol (EG) were comprehensively investigated through theoretical prediction and experimental verification. The proton capture ability of the anionic ligand from EG is positively correlated with the activity of the catalyst, as the generation of EG- was crucial for the chain breaking reaction via nucleophilic attack on the carbonyl group. Furthermore, density functional theory calculations demonstrated that the HTBD cation and anionic ligands work in a cooperative manner in the PET glycolysis reaction, where the ligands abstract a proton from EG to generate EG- and provide a proton to produce the bis(hydroxyalkyl)terephthalate (BHET) product. The rate-determining step is the nucleophilic attack step, where the Gibbs energy barriers (ΔG≠) increase in the order of 29.7 kcal mol-1 (HTBD-OAc) < 30.2 kcal mol-1 (HTBD-CH3CH2COO) < 31.4 kcal mol-1 (HTBD-HCOO) < 35.7 kcal mol-1 (HTBD-CH3COCOO) < 36.9 kcal mol-1 (HTBD-NO3). This is confirmed from the experimental results that the BHET yields decrease in the order of 84.8% (HTBD-OAc) > 82.4% (HTBD-CH3CH2COO) > 80.2% (HTBD-HCOO) > 73.6% (HTBD-CH3COCOO) > 4.7% (HTBD-NO3). These findings offer valuable guidance for designing more efficient metal-free protic ionic salts, promoting sustainable PET recycling.

New Search Method for Sbox-Related Impossible Differentials

The impossible differential attack is one of the most fundamental tools of cryptanalysis and has been successfully applied to a large variety of block ciphers. In a typical impossible differential attack, the foundation and first step is to construct an impossible differential. Nowadays, two kinds of most commonly used approaches in impossible differential construction are the matrix-based and tool-aided automatic search methods. In this paper, we proposed a new search method combining an early-abort strategy with the guess-and-determine technique to find longer impossible differentials. Compared with the previous matrix-based search methods, ours has taken the details of Sbox into consideration, while compared with the tool-aided methods, ours is independent of a third-party solver and is applicable to ciphers with large (≥8 bits) Sboxes, which could be tough work for tool-aided methods to cover. Therefore, more accurate results could be obtained. To prove the effectiveness of our method, it is applied to CSA and FOX64. For CSA, 23/24/25-round impossible differentials were found for CSA-BC. This improved the longest impossible differential distinguisher so far by 1/2/3 rounds. And 25-round key recovery attacks are performed against CSA-BC and CSA, which are the longest impossible differential attack by now. For FOX64, we proved that an impossible differential for its round function always implies an impossible differential for 4-round FOX64, and new types of impossible differentials were able to be found. Our method has its own advantage in dealing with large Sboxes, and it may be a possible direction to search for Sbox-related impossible differentials.

CMUA-Watermark: A Cross-Model Universal Adversarial Watermark for Combating Deepfakes

Malicious applications of deepfakes (i.e., technologies generating target facial attributes or entire faces from facial images) have posed a huge threat to individuals' reputation and security. To mitigate these threats, recent studies have proposed adversarial watermarks to combat deepfake models, leading them to generate distorted outputs. Despite achieving impressive results, these adversarial watermarks have low image-level and model-level transferability, meaning that they can protect only one facial image from one specific deepfake model. To address these issues, we propose a novel solution that can generate a Cross-Model Universal Adversarial Watermark (CMUA-Watermark), protecting a large number of facial images from multiple deepfake models. Specifically, we begin by proposing a cross-model universal attack pipeline that attacks multiple deepfake models iteratively. Then, we design a two-level perturbation fusion strategy to alleviate the conflict between the adversarial watermarks generated by different facial images and models. Moreover, we address the key problem in cross-model optimization with a heuristic approach to automatically find the suitable attack step sizes for different models, further weakening the model-level conflict. Finally, we introduce a more reasonable and comprehensive evaluation method to fully test the proposed method and compare it with existing ones. Extensive experimental results demonstrate that the proposed CMUA-Watermark can effectively distort the fake facial images generated by multiple deepfake models while achieving a better performance than existing methods. Our code is available at https://github.com/VDIGPKU/CMUA-Watermark.

Theoretical Study of Mechanism and Product Selectivity of Metal-Catalyzed Reactions of Alkynyl Thioethers with Isoxazoles/Anthranils

Density functional theory (DFT) calculations were performed to investigate the detailed mechanisms of metal-catalyzed reactions of alkynyl thioethers with isoxazoles/anthranils. The origins of the product selectivities were also explored. The results show that when [IPrAu(CH3CN)]+ was used as the catalyst, β-selective addition gave a sulfenylated pyrrole/indole as the primary product. However, the major product was a sulfur-substituted β-keto enamide, which was formed via α-addition, when [Zn(OTf)]+ was used. The product selectivity was determined by the first nucleophilic attack step. We found that the catalyst and substrate both played crucial roles in guiding the product selectivity. The electronic effects of ligands on the catalyst and substituents on the substrate were the dominant factor. For the anthranil substrate, the preference for C1–C6 bond coupling rather than C1–C4 bond coupling can be attributed to effective conjugation between the carbonyl group and the phenyl moiety during C1–C6 bond coupling. These insights into the mechanisms of metal-catalyzed reactions of alkynyl thioethers with isoxazoles/anthranils will facilitate the design of new catalytic reactions of isoxazoles with electron-deficient alkynes.

Malware on Internet of UAVs Detection Combining String Matching and Fourier Transformation

Advanced persistent threat (APT), with intense penetration, long duration, and high customization, has become one of the most grievous threats to cybersecurity. Furthermore, the design and development of Internet-of-Things (IoT) devices often do not focus on security, leading APT to extend to IoT, such as the Internet of emerging unmanned aerial vehicles (UAVs). Whether malware with attack payload can be successfully implanted into UAVs or not is the key to APT on the Internet of UAVs. APT malware on UAVs establishes communication with the command and control (C&C) server to achieve remote control for UAVs-aware information stealing. Existing effective methods detect malware by analyzing malicious behaviors generated during C&C communication. However, APT malware usually adopts a low-traffic attack mode, a large amount of normal traffic is mixed in each attack step, to avoid virus checking and killing. Therefore, it is difficult for traditional malware detection methods to discover APT malware on UAVs that carry weak abnormal signals. Fortunately, we found that most APT attacks use domain name system (DNS) to locate C&C server of malware for information transmission periodically. This behavior will leave some records in the network flow and DNS logs, which provides us with an opportunity to identify infected internal UAVs and external malicious domain names. This article proposes an APT malware on the Internet of UAVs detection method combining string matching and Fourier transformation based on DNS traffic, which is able to handle encrypted and obfuscated traffic due to packet payloads independence. We preprocessed the collected network traffic by converting DNS timestamps of DNS request to strings and used the trained random forest model to discover APT malware domain names based on features extracted through string-matching-based periodicity detection and Fourier transformation-based periodicity detection. The proposed method has been evaluated on the data set, including part of normal domains from the normal traffic and malicious domains marked by security experts from APT malware traffic. Experimental results have shown that our proposed detection method can achieve the accuracy of 94%, which is better than the periodicity detection algorithm alone. Moreover, the proposed method does not need to set the confidence to filter the periodicity with high confidence.

Adversarial attacks through architectures and spectra in face recognition

• We propose a novel adversarial attack on face images captured in different spectra. • The attacks are transposed across different DNN architectures and different spectra. • The attacks are composed by a FGSM attack step and a noise transfer step. • The attacks are performed on two different dataset: VIS-NIR and ViS-TH. • The results show that the DNN architectures have different responses to the attack. The ability of Deep Neural Networks (DNNs) to make fast predictions with high accuracy made them very popular in real-time applications. DNNs are nowadays in use for secure access to services or mobile devices . However, as DNNs use increased, at the same time attack techniques are born to “break” them. This paper presents a particular way to fool DNNs by moving from one spectrum to another one. The application field we explore is face recognition. The attack is first built on a trained Face DNN on Visible, Near Infrared or Thermal images, then transposed to another spectrum to fool another DNN. The attacks performed are based on the Fast Gradient Sign Method with the aim to misclassify the subject knowing the DNN to attack (White-Box Attack) but without knowing the DNN on which the attack will be transposed (Black-Box Attack). Results show that this cross-spectral attack is able to fool the most popular DNN architectures. In worst cases the DNN becomes useless to perform face recognition after the attack.

Quantitative model checking for assessing the energy impact of a MITM attack on EPONs

Broadcasting the downstream traffic makes the Ethernet Passive Optical Networks (EPONs) vulnerable to eavesdropping which is typically the initial step of an active attack, such as Man in the Middle attack (MITM). A MITM attack in such a network can be implemented by employing a fake Optical Line Terminal (OLT) and overwhelm computational, bandwidth, or energy resources. The latter is of great interest since Passive Optical Networks (PONs) are estimated to be the largest energy consumers among optical networks. In this paper we use formal analysis to quantitatively assess the impact of a fake OLT attacking an EPON energy‐efficiency mechanism. Formal verification techniques, such as model checking, constitute the perfect candidate for security verification, since they can analyze systems and protocols based on rigorous model definitions. Our results show that a MITM attack increases the energy expenditure, since it enforces the Optical Network Units (ONUs) to stay in the active mode even in cases of none network traffic.

Electrochemical Ring Expansion to Synthesize Medium-Sized Lactams Through C–C Bond Cleavage

Medium-sized nitrogen heterocycles are prevalent motifs in many kinds of bioactive molecules and natural products. Owing to the unfavorable enthalpic and entropic barriers during the transition sta...

Automatic Port Scanner

In a computer network, an attack is an attempt to destroy or steal unauthorized information or make use of information as an asset. One of the attacks is a reconnaissance attack considered as the first step of a computer attack. This type of attack is mostly done by a black hat, an expert in the programmer, by scanning the internal network devices and gather vulnerability information. In this paper, shows the identification of open ports and services through the network and available IP on the network are possible to attack.

NSAPs: A novel scheme for network security state assessment and attack prediction

With the increasing complexity and scale of networks, the computer attacks increase year by year and becomes more complicated. The defenders not only need to detect malicious activity through a large number of alerts generated by intrusion detection system, but also need to use these alerts to assess security state and predict attack, so as to take proactive response to reduce the damage of cyber-attacks. In this process, it is necessary to preprocess the huge amounts of raw alerts to get the appropriate granularity, so as to improve the accuracy of the subsequent assessment and prediction model. At the same time, the security evaluation model needs to have a good explainability and comprehensive attack prediction ability, including attack event and attack step prediction, in order to provide a better decision reference for proactive response. In addition, the model should be able to adapt to zero-day attacks. To address these issues, in this paper, we propose NSAPs, a novel scheme for network security state assessment and attack prediction. First, we extract attack steps based on quantitative alert quality to reduce the amount of data. Second, we extract attack events with medium granularity from attack steps based on Semi-Markov Conditional Random Fields (semi-CRFs). The semi-CRFs can use as much alert information as possible to correlate alerts and can also take advantage of the contextual information between the attack events. Therefore, the NSAPs can provide a comprehensive attack prediction with a good explainability. Third, the extracted attack events are used as the input of the Hidden Markov Model (HMM) to assess security state. At the same time, we propose a HMM matching method based on the longest common subsequence of the attack events which makes the model adapt to the unknown alters well. Finally, we combine probability values from semi-CRFs and HMM to predict attacks. Our evaluation results indicate that the assessment and prediction of proposed scheme are more accurate and comprehensive compared with existing approaches.

Multifactor Violator Model. Condition-factors and Resource-factors

Attack scenarios with limitations were investigated. Resource-factors and condition-factors were set as two types of limitations. Resource-factors are spent at each step of attack and can be replenished completely or partially if a given attack step was successful. A situation, where successful completion of current step with one or more preceding ones is required to replenish resource-factors, is possible. After each step of attack the violator can “exchange” resource-factors to accumulate the required amount of those factors for the next step. The lack of the required amount of resource-factors may either forcefully interrupt an attack or to lower success probability or reduce the time required by protection side to discover the consequences of an attack. This article doesn’t consider the change of relative cost of resource-factors, that is caused by urgency, so that all resource-factors have fixed cost regardless of violator’s reserve of these resource-factors. &#x0D; Conditions-factors are fixed limitations for conducting an attack. Discrepancy of condition-factors makes it impossible to either start an attack or to finish the current attack’s step. In certain cases the lack in one condition-factor can be compensated with excess of another condition-factor or via spending additional resource-factors.&#x0D; The influence on resource-factors and condition-factors is laid as a basis of protection strategies. The strategy of increasing the values of condition-factors for violator decreases the total amount of attacks on a system by screening beginner violators. The threat level from groups of violators and from experienced violators will remain unchanged. The strategy of increasing the rate of resource-factors spending is designed to interrupt attacks in progress. Strategy of decreasing the amount of resource-factors that can be replenished after successful completion of certain steps of attack scenario is meant to decrease violator’s interest in attacking specifically our system and to decrease the chances of attack repetition if an attack occurred.

Real-Time Multistep Attack Prediction Based on Hidden Markov Models

A novel method based on the Hidden Markov Model is proposed to predict multistep attacks using IDS alerts. We consider the hidden states as similar phases of a particular type of attack. As a result, it can be easily adapted to multistep attacks and foresee the next steps of an attacker. To achieve this goal, a preliminary off-line training phase based on observations will be required. These observations are obtained by matching the IDS alert information with a database previously built for this purpose using a clusterization method from the CVE global database to avoid overfitting. The training model is performed using both unsupervised and supervised algorithms. Once the training is completed and probability matrices are computed, the prediction module compute the best state sequence based on the state probability for each step of the multistep attack in progress using the Viterbi and forward-backward algorithms. The training model includes the mean number of alerts and the number of alerts in progress to assist in obtaining the final attack probability. The model is analyzed for DDoS phases because it is a great problem in all Internet services. The proposed method is validated into a virtual DDoS scenario using current vulnerabilities. The results proving the system's ability to perform real-time prediction.

Comparative study of the nucleophilic attack step in the proteases catalytic activity: A theoretical study

ABSTRACT The nucleophilic attack step of the hydrolysis reaction mechanism of the glycine-glycine peptide bond mediated by the enzymatic action of various proteases was elucidated by means of DFT calculations. Five different protease models were considered; namely: cysteine (Cys), threonine (Thr), serine (Ser), aspartyl (Asp) proteases, and a metalloprotease containing zinc (Zn). The model was simplified in order to gain information about the nucleophilic attack in this type of reaction. As a comparative study, this work is focused on the trend in the reactivity of the models. According to the computed activation energies, the reactivity order was determined as follows Cys < Thr < Ser < Zn < Asp, being in all cases faster than the uncatalysed spontaneous hydrolysis. A further analysis of the reactions by means of the reaction force approach showed that the structural changes accounts for 65–90% of the total activation energy. Moreover, a natural bond orbital analysis allows the reactions to be classified as synchronous with a late transition state for all cases. Systems analogous to the Cys-protease can be proposed as a promising candidate for the design of mimetic systems capable to cleavage amide bonds.