Currently, static analysis is insufficient to deal with Android malware that employs advanced evasion techniques such as code obfuscation and dynamic loading. Therefore, hybrid analysis that combines static structure and dynamic behavior has become the mainstream trend. However, existing hybrid analysis methods often adopt simple feature concatenation or shallow fusion mechanisms, which cannot effectively integrate heterogeneous static and dynamic features or capture the complex correlations between structure and behavior. To address this, we propose a hybrid heterogeneous graph-based Android malware detection method via multi-evidence similarity fusion, named HHGDroid. The function call graph generated by static analysis and the event graph obtained through dynamic analysis are connected through a comprehensive similarity of multiple evidences such as semantics, permissions, and time frequency, ultimately forming the hybrid heterogeneous graph with multiple heterogeneous nodes and edges. Our constructed hybrid heterogeneous graph is the first one that simultaneously possesses static and dynamic features. Finally, we improve Reliability-Calibrated Heterogeneous Graph Transformer (RCHGT) to learn the multiple relationships in the hybrid heterogeneous graph, which can automatically distinguish reliable and unreliable edges during the information propagation stage. We conduct experiments on real Android malware applications and achieved an F1-score of 97.87%, outperforming the state-of-the-art methods. Additionally, we verify our method on an unknown malware dataset and obtained an F1-score of 81.52%, which is superior to existing methods. HHGDroid is a novel and effective method for detecting Android malware.

Learning-based methods have been widely applied in the field of Android malware detection. However, adversarial samples pose a serious challenge to such methods, as carefully constructed adversarial samples may evade detection by these detectors. To evaluate the robustness of the mainstream Android malware detection, in this paper, we propose a novel differentiated adversarial perturbation generation method in problem space. We first slice a large number of benign applications to get a set of code slices that preserve context semantics. An improved optimal perturbation screening method based on Hierarchical Attention Network is proposed to effectively select the optimal slice from the code slice set as the perturbation of the target attack model. We perform dynamic adaptive compute based on the target attack model to achieve the optimal adversarial perturbation. After adding perturbation to the target sample, the sample is repackaged and signed to verify the adversarial effect of the detection model. The experimental results on multiple malware datasets show that the adversarial samples generated by our method can significantly reduce the accuracy of the target detectors and achieve better adversarial attack effect compared with the existing methods.

Robotic-assisted laparoscopic surgery has made progress in addressing many of the technical challenges associated with conventional laparoscopy. Recently, a new robotic surgical system, Carina Platform, has been developed. The aim of this study is to evaluate the feasibility and safety of Carina for proctectomy surgery in preclinical models. Port and cart positions were initially determined by surgical simulation based on three-dimensional reconstruction. Six pigs, divided equally into acute and chronic groups, underwent robotic-assisted proctectomy using the Carina robotic system. Operative time, device errors and perioperative complications were recorded. All animals were observed postoperatively to assess their mental status, and after four weeks, the chronic pigs were euthanized for necropsy to evaluate the recovery of the anastomotic stoma. The optimal port and patient cart positions were further assessed in three human cadavers. Robotic-assisted proctectomy was successfully completed in all porcine subjects. No device-related intraoperative complications were observed. An autopsy examination revealed that the chronic pigs, which had been followed up for 28 days, had made a favorable recovery. The configurations of ports and patient carts were successfully validated in human cadaver models that could simulate the completion of a rectal resection. The feasibility and safety of the newly developed Carina robotic system for performing proctectomy were successfully demonstrated in porcine and cadaveric models. Further studies are required to validate its clinical application in human patients.

Context: In recent years, the rapid development of deep learning technologies has opened new avenues for vulnerability detection. Many existing methods convert source code into images for detection but often overlook the quality of the generated images, resulting in the loss of critical vulnerability features. Objective: This study aims to address the challenges of vulnerability feature loss and image quality issues in vulnerability detection. Specifically, we investigate the effectiveness of a novel pixel-row oversampling method based on code line concatenation for generating enhanced code images. Methods: We propose a new system, CIPRO, which leverages pixel-row oversampling to integrate structural and semantic information from code. CIPRO generates interpolated code lines, creating continuous and feature-rich images for improved vulnerability detection. We validate this method through theoretical analysis and empirical studies on three datasets: Devign, Reveal, and SARD. Results: Our experiments show that CIPRO achieves significant performance improvements, with F1-score increases of 5.36%, 19.62%, and 6.46% on Devign, Reveal, and SARD datasets, respectively, compared to existing methods like VulCNN. CIPRO also demonstrates high time efficiency and strong visualization capabilities, meeting practical large-scale requirements. Conclusion: CIPRO's pixel-row oversampling method addresses the shortcomings of traditional code image generation methods, providing a scalable and effective solution for vulnerability detection.

Android dominates the global mobile ecosystem, while malware poses a serious security threat to ordinary users. Previous research has focused more on the bytecode features of target Android malware and less on analyzing features from various other sources such as resource files and native code. Furthermore, single-modal features lack the utilization of feature files after reverse engineering of applications. To address these limitations, we propose a Residual-Gated Multimodal Network (RGMN) that adaptively fuses configuration text with grayscale image features derived from binary byte streams for Android malware detection. First, we extract basic configuration text features such as permissions and configurations from AndroidManifest.xml. Then, our framework extracts underlying code logic, resource configurations, and auxiliary runtime components by mapping binary data from classes.dex, resource files, and .so files into three-channel grayscale image representations. Compared with previous fusion strategies, our multimodal feature fusion strategy can better resolve modal conflicts between configuration text and binary byte stream images. Our multimodal feature fusion method does not require complex dynamic and static code analysis, making it applicable to a wider range of scenarios.

Liver metastasis is the primary cause of mortality in colorectal cancer (CRC) patients. To decipher the underlying mechanisms, we performed single-cell RNA sequencing (scRNA-seq) on paired primary colorectal tumors, adjacent tissues and liver metastases from three CRC liver metastasis (CRLM) patients, alongside colorectal tumors and adjacent tissues from three non-metastatic CRC patients. Our analysis revealed a significant enrichment of Enolase 2-expressing (ENO2⁺) cancer cells in CRLM patients compared to their non-metastatic counterparts. Functional characterization, supported by bioinformatics and murine models, demonstrated that ENO2⁺ cancer cells exhibit enhanced epithelial-mesenchymal transition (EMT) and are critical drivers of CRLM. Mechanistically, the ENO2 protein directly binds to macrophage migration inhibitory factor (MIF) within cancer cells, stabilizing MIF by inhibiting its C-terminus of Hsc70-Interacting Protein (CHIP)-mediated ubiquitination and degradation. This ENO2-MIF interaction activates MIF signaling, fostering robust tumor cell-macrophage crosstalk that promotes M2 macrophage polarization, which is validated by spatial transcriptomics showing the colocalization of ENO2⁺ cancer cells and M2 macrophages. Crucially, both organoid and in vivo models confirmed that ENO2 in CRC cells is essential for inducing M2 macrophage polarization via the MIF pathway, thereby facilitating liver metastasis. Knockout of ENO2 significantly suppressed tumor growth and liver metastasis in mouse models. An inhibitor of the ENO2-MIF interaction, pyrithioxin, can effectively reduce the burden of liver metastasis in mice. Collectively, our findings identify ENO2 as a key driver of CRLM by stabilizing MIF to orchestrate M2 macrophage polarization, highlighting the ENO2-MIF axis as a promising therapeutic strategy for CRLM.

Neoadjuvant therapy has become a cornerstone in the management of locally advanced rectal cancer (LARC). In this single-arm, open-label phase II study, we evaluated the efficacy and safety of total neoadjuvant chemotherapy (TNT) using a CapOX regimen combined with a programmed cell death protein 1 (PD‑1) antibody (sintilimab) and interleukin‑2 (IL‑2) in patients with microsatellite stable (MSS)/defining proficient mismatch repair (pMMR) LARC. A total of 33 patients, aged 18-75 years, with rectal tumors located within 12 cm from the anal verge and staged as cT3/4N_anyM0 or cT_anyN⁺M0, were enrolled. Patients received a regimen consisting of oxaliplatin, sintilimab, capecitabine, and IL‑2 administered in a three‑week cycle, with response evaluations performed after every two cycles. Following six cycles of treatment, 33 patients underwent radical surgery, achieving a 100% R0 resection rate. The pathological complete response (pCR) rate was 42.4% (95% CI: 25.68-59.16%) while the remaining 19 patients (57.6%, 95% CI: 41.07-74.09%) were assessed as having a partial response. The tumor regression grades (TRG) were, TRG1: 3 cases (9.1%, 95% CI: 1.90-25.97%), TRG2: 14 cases (42.4%, 95% CI: 26.27-60.38%), and TRG3: 2 cases (6.1%, 95% CI: 0.73-20.37%), respectively. Surgical safety reported as no cases of grade B/C anastomotic leakage or bowel obstruction. Adverse events (AEs) were manageable, with a 21.2% incidence of grade 3 events and no grade IV/V events or treatment‑related deaths. With a median follow‑up of 25.5 weeks, no recurrences were observed. Analysis of blood and tissue samples from patients with different treatment outcomes revealed significant activation of CD8+ T cells, NK cells, and M1 macrophage subsets in the tumor microenvironment of patients achieving pCR. These compelling data demonstrate promising efficacy with a favorable safety profile in MSS/pMMR LARC. These findings warrant studies to validate this regimen as a novel treatment paradigm for rectal cancer. ClinicalTrials.gov registration: NCT06108596.

Colorectal cancer (CRC) remains a leading cause of cancer mortality, and the molecular drivers of progression and distant metastasis are incompletely understood. By integrating differential expression and survival analyses of TCGA CRC cohorts with HOX family genes, we identified HOXC11 as a key metastasis-associated factor. HOXC11 was markedly upregulated in CRC tissues and cell lines, with higher expression in metastatic lesions than in primary tumors, and elevated HOXC11 correlated with poor patient prognosis. HOXC11 functionally increased CRC cell proliferation, migration, and invasion in vitro and facilitated tumor growth and metastasis in vivo. Mechanistically, HOXC11 directly bound to the CAMK2A promoter and transactivated CAMK2A, leading to increased phosphorylated CAMK2A and initiation of the NF-κB pathway, which facilitated p65 nuclear translocation and induced CXCL5 expression to drive CRC progression. Conversely, CXCL5 signaling through CXCR2 upregulated HOXC11 via the ERK1/2-SP1 axis, forming a positive feedback loop. Notably, combined inhibition of CAMK2A (KN-93) and CXCR2 (SB265610) significantly attenuated HOXC11-mediated proliferation and metastasis. Collectively, these findings define a HOXC11-CAMK2A-NF-κB-CXCL5 circuit as a potential therapeutic target in CRC.

Colorectal cancer liver metastases (CRLM) present significant treatment challenges, requiring multimodal conversion therapies. Identifying factors that influence treatment outcomes is crucial for improving clinical management. This retrospective cohort study included 286 patients with synchronous CRLM who underwent conversion therapies on the basis of sequencing results. Patients were categorized into successful conversion therapy group (SCTG) and failed conversion therapy group (FCTG). Clinical factors and genomic mutations were analyzed for associations with therapy outcomes and survival. Among the patients, 106 (37.1

Emerging evidence highlights the role of SCF E3 ligases, consisting of SKP1, cullin-1, and F-box proteins, in cancer biology by regulating the ubiquitination and degradation of key proteins. This study identifies F-box only protein 44 (FBXO44) as an oncogene in colorectal cancer (CRC). FBXO44 is upregulated in CRC patients and correlates with poor prognosis. Knockdown of FBXO44 inhibits CRC cell proliferation and organoid growth, as well as xenograft tumor growth and AOM/DSS-induced intestinal tumorigenesis. Conversely, FBXO44 overexpression accelerates tumor growth in vitro and in vivo. Mechanistically, FBXO44 targets Forkhead box protein P1 (FOXP1) for degradation. Aurora kinase A (AURKA) phosphorylates FOXP1 at Ser440, enhancing FBXO44 binding, leading to K48-linked ubiquitination at K377 and proteasomal degradation. This degradation relieves FOXP1 repression of Cyclin E2, promoting CRC cell proliferation. In summary, FBXO44 is an oncogene that promotes CRC tumorigenesis by degrading FOXP1 and upregulating Cyclin E2, offering a potential therapeutic target for CRC.

Colorectal cancer (CRC) progression and metastasis involve numerous regulatory factors. Among these, cellular retinoic acid-binding protein 2 (CRABP2) has been implicated as both a tumor activator and suppressor. Here, it is aimed to clarify the role of CRABP2 in CRC growth and metastasis and explore the underlying molecular mechanisms mediating its cellular functions. Using both in vitro and in vivo models, including a colonocyte-specific CRABP2 conditional knockout mouse model (Crabp2ΔIEC) and a subcutaneous tumorigenesis assay in BALB/c nude mice, it is shown that nuclear CRABP2 enhances tumor growth by interacting with and downregulating the tumor suppressor RB1, whereas cytoplasmic CRABP2 suppresses CRC liver metastasis by interacting with AFG3L2 and promoting mitophagy. In addition, the AFG3L2-SLC25A39 axis is identified as a distinct mechanism by which cytoplasmic CRABP2 increases mitochondrial glutathione stability to promote cell proliferation independent of the nuclear RB1 pathway. Notably, analysis of tissue from CRC patients reveals that CRABP2 protein has distinct prognostic implications and functional roles in the progression and metastasis of CRC dependent on its subcellular localization. Ultimately, by elucidating the role of CRABP2 in CRC, it is aimed to provide new insight into disease pathogenesis and inform the development of therapeutic interventions.

Android has occupied an important share of the operating system of intelligent terminal devices in the Internet of Things (IoT), and the malicious applications of Android have increased rapidly, posing a serious threat to the security of IoT. Machine learning has advanced significantly in the detection of android malware. In order to protect intellectual property, Android developers have begun to use packing techniques to enhance the security of their applications. However, attackers can also pack their malware, which may make feature extraction ineffective and interfere with the prediction results of learning-based classifiers. For this issue, we have designed and implemented a tool by dynamically loading the original DEX using a shell DexClassLoader to generate a batch of packed Android applications. And we have verified that several existing methods fail when faced with packed samples. Therefore, we propose a novel malware detection method called DTDroid that can resist code packing. DTDroid automatically captures network traffic characteristics of target samples based on fuzzy testing and network traffic packet extraction. At the same time, the dynamic behavior characteristics of the target application can be obtained by monitoring the corresponding runtime function calls and system status. The extracted two types of features are contextually spliced and converted into grayscale images, and then detected based on deep learning model. Experimental results show that the detection accuracy of our method reaches 94.22% and 95.14% respectively on two kinds of packed datasets, indicating that DTDroid has better robustness for packed samples than the existing methods.

PJA2 is documented to degrade various substrates. Nevertheless, the role of PJA2 as an E3 ubiquitin-protein ligase in colorectal cancer (CRC) progression remains unexplored. The correlation between PJA2 mRNA levels and clinical characteristics is investigated using data from The Cancer Genome Atlas (TCGA) database. Quantitative real-time polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC) are utilized to evaluate PJA2 expression levels in CRC tissues. The biological functions of PJA2 are confirmed through colony formation assays and azoxymethane/dextran sulfate sodium (AOM/DSS) mouse model of CRC, among other experimental approaches. The underlying molecular mechanisms of PJA2 action are elucidated using RNA sequencing (RNA-seq), co-immunoprecipitation (co-IP), proximity ligation assay (PLA), and chromatin immunoprecipitation (ChIP). Our research discovered that PJA2 is downregulated in CRC tissues and decreased PJA2 expression correlates with poor prognosis. Functionally, in vivo and in vitro experiments uncovered that PJA2 inhibits tumor cell proliferation and promotes apoptosis. Mechanistically, PJA2 recognized histone deacetylase 2 (HDAC2) via its RING-B-box domain (RBD) and bind to the N-terminal of HDAC2, facilitating ubiquitination at the lysine 90 (K90) residue. PJA2-mediated degradation of HDAC2 counteracts the transcriptional repression of the interferon-induced protein with the tetratricopeptide repeats (IFIT) family, thereby suppressing CRC progression. The data demonstrates that PJA2 suppresses CRC progression through the PJA2/HDAC2/IFIT axis, and its expression is regulated by HDAC2, thus constituting a positive feedback loop. Consequently, PJA2 may serve as a potential therapeutic target for CRC, and interrupting this feedback loop can represent a viable treatment strategy to restrain CRC progression.

Metabolic reprogramming is a notable hallmark of cancer biology, especially aerobic glycolysis. Some clinical trials attempt to target cancer metabolism to develop therapeutic agents. However, the results have been not satisfactory. Here, we report that REEP6 is significantly upregulated and promotes glycolysis and tumorigenesis in CRC. Moreover, REEP6, as a molecular scaffolder, bridges the PRMT5-PGAM1 complex, which enhances the PRMT5-mediated symmetric dimethylarginine (SDMA) of PGAM1 at R40. The methylated PGAM1 possesses dramatically enhanced enzymatic activity and therefore boosts glycolytic flux in CRC cells. More than that, our results showed that combined treatment with specific shRNA and inhibitors exhibits synergistic anti-tumor efficacy in CRC, which may shed light on the development of a promising therapy in CRC.

Dear Editor, Our study provides a comprehensive analysis of the correlation between specific somatic mutations and the lymph node metastasis in colorectal cancer (CRC) patients with T1/2 stage, addressing a significant gap in the early-stage prognosis and individualised treatment of CRC. By leveraging next-generation sequencing (NGS) and clinical data, we identified key mutations and pathological characteristics that can serve as robust predictors of lymph node metastasis, thereby enhancing clinicians' ability to stratify T1/2 CRC patients based on metastatic risk. CRC remains a prevalent malignancy worldwide, with early-stage cases, particularly T1/2 stages, generally associated with favourable outcomes.1 However, when lymph node metastasis is present, the prognosis worsens significantly.2 Lymph node involvement in early-stage CRC is a strong indicator of potential distant metastasis and recurrence, highlighting the urgency of effective preoperative risk assessment in T1/2 patients.3 Existing imaging techniques, such as enhanced computed tomography and magnetic resonance imaging, offer limited sensitivity and specificity for lymph node assessment, especially in early stages where inflammation or small metastatic nodes may escape detection.4, 5 Therefore, identifying molecular markers that can predict lymph node involvement with high accuracy is paramount to guiding treatment decisions in early-stage CRC. Our study contributes to this goal by identifying specific genetic mutations and clinical features associated with metastatic risk, thus enabling a more refined preoperative evaluation of T1/2 stage CRC patients. We conducted a retrospective cohort study including 212 T1/2 CRC patients (Figure S1), who were categorised based on lymph node involvement into T1/2N+ and T1/2N‒ groups and matched using propensity score analysis (Figure 1 and Table S1). Our NGS data showed that mutations in four genes (LRP1B, KMT2B, TSC2 and BRAF) were significantly more frequent in the T1/2N+ group (Figure 2), and the presence of any of these mutations correlated with reduced overall survival (Figure S2). Among the four, BRAF mutations have been widely recognised in literature as a poor prognostic factor in advanced CRC,6 but their impact in early stages has remained largely unexplored until now. Additionally, LRP1B, KMT2B and TSC2 mutations appear to be novel findings in the context of early-stage lymph node metastasis in CRC, indicating that they could potentially be unique markers for predicting outcomes in T1/2 patients (Figure 3). Our analysis also showed that patients with LRP1B mutations had particularly poor survival outcomes, with a median survival of 22.2 months. In addition to these molecular findings, there was no significant difference in the distribution of MSI status and TMB between the two groups; however, we observed significant differences in clinical characteristics between T1/2N+ and T1/2N‒ patients. T1/2N+ patients had higher rates of lymphovascular invasion, poor tumour differentiation and aggressive histological subtypes. The presence of lymphovascular invasion, an established risk factor for metastasis, was notably higher in T1/2N+ patients (66 .0% vs. 51.9%, p = .036). Poorly differentiated tumours were also more common in T1/2N+ patients, with the proportion being nearly double that of the T1/2N‒ group (32.1% vs. 16.1%, p = .019). These pathological characteristics underscore the heterogeneity within T1/2 stage CRC and the importance of integrating clinical and molecular data to predict metastatic risk accurately. Histological type also played a role, with higher proportions of mucinous and signet-ring cell carcinomas observed in the T1/2N+ group, both of which are histologically aggressive subtypes associated with poor prognosis (Table S2). Given the clinical importance of preoperative assessment for lymph node metastasis, we developed a predictive nomogram model incorporating the identified mutations, lymphovascular invasion, tumour differentiation and histological type (Figure 4A and Table S3). This nomogram provides a scoring system where each risk factor contributes to an overall score that correlates with the probability of lymph node metastasis. Furthermore, this model exhibits high predictive accuracy, with an area under the curve of 81.3%, underscoring its reliability in evaluating metastatic risk in early-stage CRC patients (Figure 4B). Clinicians can leverage this tool to make more informed decisions about surgical strategies, including the extent of lymph node dissection, and to assess the necessity of postoperative adjuvant therapies. By stratifying patients based on their risk scores, the nomogram allows for a more personalised treatment strategy, potentially improving survival outcomes and reducing the need for overtreatment in low-risk cases. Our study also sheds light on the prognostic value of these specific genetic mutations in early-stage CRC. For instance, BRAF mutations are already associated with poor outcomes in CRC, particularly in metastatic cases. Our study confirms this trend even in early-stage CRC, suggesting that BRAF mutations could serve as a marker for aggressive disease progression, even before metastasis is evident. Similarly, KMT2B and TSC2 mutations, although not well-studied in CRC, have been implicated in other malignancies for their roles in enhancing tumour aggression and stemness,7, 8 further supporting their inclusion as high-risk markers in our model. The mutation of LRP1B, typically associated with poor prognosis in other cancer types,9 also emerged as a significant indicator of poor survival in our cohort, underscoring the potential importance of this mutation in the early metastatic pathway of CRC. Our findings emphasise the need for a multidimensional approach in the management of T1/2 stage CRC patients. The integration of molecular and clinical markers in a predictive model provides a more accurate assessment of lymph node metastasis risk, which is crucial for guiding treatment decisions in early-stage CRC. Current clinical practice primarily reserves targeted therapies for advanced stages, but our findings suggest that high-risk T1/2 patients may also benefit from targeted approaches aimed at specific genetic mutations, which could ultimately improve their prognosis. Our study also has several limitations. First, the retrospective nature of the study design limits the level of evidence provided. Future prospective studies are needed to provide more robust and comprehensive results. Moreover, the application of this predictive model requires further precise analysis to establish specific weights, which will be crucial for guiding subsequent treatment decisions. In conclusion, our study identifies mutations in LRP1B, KMT2B, TSC2 and BRAF as significant prognostic markers in T1/2 CRC, along with critical clinical features such as lymphovascular invasion, poor tumour differentiation and aggressive histological types. By combining these factors into a predictive nomogram model, we provide a tool that enables accurate preoperative assessment of lymph node metastasis risk, facilitating a more personalised treatment approach for early-stage CRC patients. Our comprehensive approach represents a significant step towards optimising outcomes and enhancing the quality of care for T1/2 stage CRC patients. Conceptualisation, methodology, software: Priya Singh. Data curation and writing—original draft preparation: Junwei Tang. Methodology, visualisation and investigation: Yue Zhang. Supervision: Dongjian Ji. Software and validation: Lu Wang. Methodology, writing—reviewing and editing, and project administration: Han Zhuo. Methodology, software Priya Singh and project administration: Yaping Wang. Supervision, project administration and funding acquisition: Yueming Sun. We are grateful to Nanjing GENESEEQ Co. for assistance with sequencing and/or bioinformatics analysis. The authors declare they have no conflicts of interest. This work was supported by the National Natural Science Foundation (82273406, 82473049 and 82304221), Basic Research Program of Jiangsu Province (BK20221415 and BK20230730), Jiangsu Key Medical: Discipline (General Surgery; grant no. ZDXK202222) and China Postdoctoral Science Foundation (2022M721679). This cohort study was approved by institutional review boards of the First Affiliated Hospital of Nanjing Medical University. Informed consent was obtained from all patients. The data sets used in the current study are available from the corresponding author upon reasonable request. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Vulnerability detection is crucial in the field of software security. However, existing methods often suffer from interference caused by redundant information and insufficient cross-line semantic dependencies when handling large-scale and complex source code, which limits detection performance. To address these challenges, this paper proposes VULDA, a source code vulnerability detection method that integrates a Vulnerability-aware Code Mapping Graph (VCMG) with Local Dependency Context Aggregation (LDCA). VCMG significantly reduces redundancy in graph structures by aligning multi-granularity semantics to line-level nodes, thereby enhancing representational compactness. Additionally, it incorporates static heuristic rules and structural features to weight nodes, effectively improving the model’s sensitivity to key vulnerability-related code. Building upon this, the LDCA module aggregates both control-flow graph (CFG) and data-dependency graph (DDG) paths, achieving dual-context aggregation of logical and data semantics, which further enhances the model’s ability to express complex vulnerability patterns. Experimental results on multiple real-world datasets, including SARD, Reveal, and FFmpeg+Qemu, demonstrate that VULDA outperforms existing methods across various metrics, notably achieving a 23.09

Pediatric Acute Lymphoblastic Leukemia (ALL) is the most common malignant tumor of the hematological system in children, and its relapse after treatment has consistently been a significant factor hindering prognosis. This study aimed to develop a blood-based non-invasive method for predicting relapse in children with ALL. Two cohorts of pediatric ALL patients were analyzed. Through high-throughput profiling, three miRNAs and three circRNAs were identified as potential biological markers, exhibiting a gradient increase in expression from healthy controls to the relapsed group. Logistic regression analysis revealed the superior predictive ability of the combined non-coding RNA panel compared to individual groups. A nomogram incorporating the non-coding RNA panel and other clinical risk features was developed. Combining the non-coding RNA panel with relevant risk features could enhance predictive accuracy. The non-coding RNA panel remained an independent predictor of relapse in the validation cohort, and its combination with clinical features formed a superior risk stratification model. In conclusion, this blood-based non-invasive method holds promise for predicting relapse in pediatric ALL patients at the time of initial diagnosis. The non-coding RNA panel, along with clinical risk features, may significantly impact patient care and outcomes.

With the increasing number of big data applications, large amounts of valuable data are distributed in different organizations or regions. Federated Learning (FL) enables collaborative model training without sharing sensitive data and is widely used in AI medical diagnosis, economy, and autonomous driving scenarios. However, it still leaks the privacy from the gradient exchange in federated learning. What’s worse, state-of-the-art work, such as Batchcrypt, still suffers from computational overhead due to a considerable amount of computation and communication costs caused by homomorphic encryption. Therefore, we propose a novel symmetric key-based homomorphic encryption scheme, Sym-Fed. To unleash the power of symmetric encryption in federated learning, we combine random masking with symmetric encryption and keep the homomorphic property during the gradient exchange in the federated learning process. Finally, the security analysis and experimental results on real workloads show that our design achieves performance improvement 6× to 668× and reduces the communication overhead 1.2× to 107× compared with the state-of-the-art work, BatchCrypt and FATE, without model accuracy degradation and security compromise.