ZSP Patentanwälte PartG mbB

Embodiments of the present disclosure include a vertical-cavity surface-emitting laser device, the device including a mesa with a non-circular cross-section. Embodiments also include a plurality of layers with one or more oxide layers. Embodiments may also include one or more non-oxide layers. The mesa may be operable as a dielectric top mirror.

An Al-based method for detecting adversarial cell telemetry vulnerabilities in real time resulting from O-RAN's openness and multi-operator untrusted structure, comprising the steps of: extracting relevant time-series features from the reporting cells and UEs Key Performance Indicators (KPIs); training a dedicated AE model for each cell; generating new enriched feature vectors that capture contextual information from both each specific cell and the entire network using compressed latent space from each AE model, concatenated with the aggregated latent spaces from all other cells; training an additional AE for each cell using the new enriched feature vectors, thereby reconstructing the original features; training a classifier to compare the output of the second model to the input of the first model; and if the loss between the input and output exceeds a predefined threshold, classifying the input as untrusted, otherwise classifying the input as trusted.

An LLM-based method for optimizing Security Information and Event Management (SIEM)'s rule sets, comprising: using LLMs to analyze contextual components of SIEM rules in the rule sets; generating SIEM rule embeddings and using a similarity matching analysis, to identify the most similar rules in the rule sets; analyzing the most similar rules, by the LLMs, to determine rule similarity by performing semantic and syntactic comparisons of SIEM rules in the rule sets; using the comparisons for identifying and eliminating of redundant rules; and providing optimized recommendations for refining the rule set, based on the elimination of redundant rules.

An optical signal generation system for optical signal generation comprises a plurality of light sources; a plurality of optical modulators each one corresponding to one of the plurality of light sources; and a synchronized driver coupled to each of the modulators. Each light source may input a continuous wave (CW) to each optical modulator. The synchronized driver may provide a non-return-to-zero (NRZ) driver signal to each modulator and the optical modulators may modulate each received optical source signal based on the NRZ driver signal communicated to each optical modulator. Each modulated optical signal may be coupled into a transmission medium, thereby generating a multi-level pulse amplitude modulated (PAM) equivalent signal.

A method for detecting cyberattacks on satellite systems, according to which native telemetry data produced by subsystems of the satellite are collected and transmitted to the ground station for monitoring and control. Data generated by a trusted power monitoring device is also collected, for monitoring the energy consumption data from the satellite's primary power source. The satellite's energy consumption data is then transmitted to a ground station, via a separate dedicated out-of-band communication channel. The native telemetry data which is collected by a data collection apparatus, is compared (by a computerized device having at least one processor and associated memory) to the energy consumption data, collected by the trusted power monitoring device and the comparison results are analyzed by a trained ML model (such as a deep-learning-based model) on the computerized device, to detect discrepancies or anomalies being indicative of possible telemetry data manipulation. Upon detecting discrepancies or anomalies by the model, an alert regarding a potential cybersecurity event is issued by the computerized device.

An LLM-based method for detecting Advanced Persistent Threat (APT) and providing natural language interpretability of the detection findings, according to which unsupervised learning is performed, for statistical anomaly detection in provenance data of provenance-based systems. Large Language Models (LLMs) with implicit knowledge are used to detect unknown attack patterns by performing context-aware graph-based analysis of detected anomalies in the provenance data, for accurate interpretation of the detected anomalies and underlying context of potential attacks; natural language explanations, drawing connections between events, and outputting insights, based on the interpretation are provided.

An RL-based framework for energy efficient onboard anomaly detection of a satellite moving in a predetermined orbit, comprising a repository of pre-trained anomaly detection models, trained on a specific subset of samples, or rule-based detectors; onboard sensors for continuously monitoring the satellite systems and data network and the onboard collected data; an RL-based agent for deciding which model to activate each time and what features to monitor via a subset of sensors, according to the location, cyber vulnerability level, previous anomalous activity, functions to be executed and energy state of the battery of the satellite, to achieve in each state the highest performance of each anomaly detection task, while investing the lowest amount of energy possible; means for providing alerts in the event of a detected cyber-attack.

Systems and methods are provided for ring aperture vertical-cavity surface-emitting laser (VCSEL) devices for mode control. A VCSEL is provided where the optical mode and the current confinement is defined by patterning the same layer of the resonant cavity within the epitaxial layer with a ring-shaped aperture. The ring width may be small enough to be single mode along the radius. The two concentric boundaries forming the ring aperture may be of any arbitrary shape as long as the width is small enough to be single mode along the radial direction.

The present invention relates to combinations, pharmaceutical compositions and fusion molecules comprising an FGF21 (fibroblast growth factor 21) compound and a GLP-1R (glucagon-like peptide-1 receptor) agonist with optimized GLP-1R agonist / FGF21 compound activity ratio. It further relates to their use as medicaments, in particular for the treatment of obesity, being overweight, metabolic syndrome, diabetes mellitus, diabetic retinopathy, hyperglycemia, dyslipidemia, Non-Alcoholic SteatoHepatitis (NASH) and/or atherosclerosis.

The present invention provides a VWF fragment comprising the D' domain and D3 domain of VWF, a chimeric protein comprising the VWF fragment and a heterologous moiety, or a chimeric protein comprising the VWF fragment and a FVIII protein and methods of using the same. A polypeptide chain comprising a VWF fragment of the invention binds to or is associated with a polypeptide chain comprising a FVIII protein and the polypeptide chain comprising the VWF fragment can prevent or inhibit binding of endogenous VWF to the FVIII protein. By preventing or inhibiting binding of endogenous VWF to the FVIII, which is a half-life limiting factor for FVIII, the VWF fragment can induce extension of half-life of the FVIII protein. The invention also includes nucleotides, vectors, host cells, methods of using the VWF fragment, or the chimeric proteins.