Our Work

Measurement scripts and raw results behind the manuscript "Cover dithering and key-derived whitening for payload-indistinguishable provenance marking under a trained steganalyst".The bundle contains t…

Measurement scripts and raw results behind the manuscript "Cover dithering and key-derived whitening for payload-indistinguishable provenance marking under a trained steganalyst".The bundle contains the embedding rules and the detectors that read them: the deployed block-replacement channel, LSB matching, cover dithering, HILL content-adaptive embedding, and the proposed rule that replaces bit 0 of every channel with a keyed bit and whitens the payload with a keystream derived over the deployment's existing elliptic-curve key pair. It also carries the structural steganalysis over 150 operational covers, the trained steganalyser with its positive, leak and necessity controls, and the node-side timing, energy and survey files of the companion study on the same testbed.README.md maps every claim in the paper to the file that produces it. Every figure can be rebuilt from the archived CSVs without a GPU; the trained-detector rows need one, and their recorded outputs are included so the numbers can be checked without repeating the training.Model checkpoints and the image corpora are not redistributed; README.md explains why and where to obtain them.Version 4.0 is a second measurement pass, made after an internal review found five defects in the first. The embedding protocol is corrected: the payload length is a protocol constant, so the receiver no longer needs a value that the offset it locates was supposed to carry, and the key dictionary holds the nonce alone. The dithering stream is derived from the shared secret under a third label instead of from an unspecified generator, which is what the security argument assumes. Experiment key pairs are seeded, because unseeded ones gave a single run three different shared secrets once the module was re-imported in each data-loader worker. The null arms are rerun at the full budget with early stopping disabled and their per-epoch loss recorded: the trained detector measures AUC 0.4999, 0.5002, 0.5000, 0.4999 and 0.5000 under five seeds with the loss at ln 2 throughout, while removing the whitening returns it to 0.7335.This version also adds the two comparisons the paper most needed. HILL at the same payload costs 83.13 dB PSNR and changes 166 channels, against 77.24 dB and 638 for the deployed scheme and 51.14 dB and 259,643 for the proposed rule; the trained detector separates HILL at 0.6048, 0.5543, 0.6175, 0.5612 and 0.6072 under five seeds (mean 0.589), four of the five still rising when the budget ended. And the detector itself is calibrated against the field's benchmark for the first time: on BOSSBase 1.01 with HILL at 0.4 bits per pixel the same architecture reaches AUC 0.7499 and 66.9 per cent accuracy, below what the literature reports for full-size networks of this family, which is the honest reading of a detector with eight residual kernels rather than thirty.Three further measurements complete this version. The residual of the unwhitened arm is apportioned: with the constant length header randomised in every frame and base64 kept, the detector still reaches AUC 0.7115; with the header kept and the base64 part replaced by uniform bits it stays at 0.5000, so base64's period-8 structure carries the residual and the header carries nothing the detector learns. Dithering itself is visible: raw frames against the same frames after keyed dithering, with no payload in either, separate at AUC 1.0000. At the operational payload on BOSSBase 1.01, whose covers were never JPEG compressed, the deployed scheme is separated at 0.7493 and the proposed rule stays at 0.5000 with the loss at ln 2 throughout.Version 4.1 adds three training seeds to each of the two arms whose result a pair of seeds could not settle: the proposed rule stays at chance under all five, with the loss at ln 2 from the second epoch on, and HILL is separated under all five. It also updates the README to the section numbering and title of the revised manuscript, and stores every training log as UTF-8.

ElectromagneticFields v0.9.1 Diff since v0.9.0 Changed The package requires GeometricBase 0.15.0 and Julia 1.11. The [compat] floors rise from GeometricBase = "0.14" and julia = "1.10&…

ElectromagneticFields v0.9.1 Diff since v0.9.0 Changed The package requires GeometricBase 0.15.0 and Julia 1.11. The [compat] floors rise from GeometricBase = "0.14" and julia = "1.10", because GeometricBase 0.15 declares its stubs public and requires Julia 1.11. Julia 1.10 users keep 0.9.0. The test suite follows the tree's test convention, with a JET check of the generated field accessors. The [compat] entries of the test and docs environments for dependencies of the package are removed. Nothing under src/ changes for these. Fixed Two equilibria of one type that name different parameters no longer share generated code. The key of the field cache did not carry the parameter names, so a second instance with other names but the same parameter shapes got the first one's code and read its parameters in the wrong slots, with no error. The key now carries parameter_names. No shipped equilibrium defines get_parameters, so none was affected. The documentation of the perpendicular frame a, c now says where it is discontinuous. The generated code is unchanged. Merged pull requests: Document where the perpendicular frame is discontinuous (#23) (@michakraus) Key the field cache on the parameter names (#24) (@michakraus) Move the test suite to the test convention (part M18) (#25) (@michakraus) Tests: JET optimisation analysis of the field accessors (part J5) (#26) (@michakraus) Remove the shared [compat] entries from test/ and docs/ (#27) (@michakraus) Accept the placeholder in the plot_equilibrium docstring for fatou (#28) (@michakraus) Require GeometricBase 0.15.0 and Julia 1.11, version 0.9.1 (#29) (@michakraus)

CHILD-EXIT is a conceptual and research framework developed by CIRITS to investigate how separation from child criminal exploitation and forced criminality may become a sustainable exit over time.The …

CHILD-EXIT is a conceptual and research framework developed by CIRITS to investigate how separation from child criminal exploitation and forced criminality may become a sustainable exit over time.The framework does not introduce the general notion of sustainable exit. Its specific contribution is to formulate a child-focused, longitudinal and falsifiable architecture linking protective separation, liminal transition, functional replacement, identity transition, social recognition and reincorporation, ecological and institutional continuity, and longitudinal sustainability.It proposes a set of explicit research gaps and testable propositions intended to guide systematic evidence review, longitudinal research and future empirical evaluation.The central research question is: under what conditions does separation from forced criminality become a sustainable exit for children and adolescents, and through which mechanisms is that durability produced, weakened or reversed?CHILD-EXIT is presented as a framework under development and not as a validated intervention model.

Exact Derivation of the Quarter-Quantum Sector, the Four-Cycle Reconnection Algebra, and the ν = 1/2 Ground-State Geometry from the 861-Node Simplicial Complex.

Batch upload of files from directory 'output'.

This contains all the supplemntary data of the paper "Chromosome level genome assembly and comparative analysis identify lineage-specific effector evolution in Fusarium luffae driving pathogenicity in…

This contains all the supplemntary data of the paper "Chromosome level genome assembly and comparative analysis identify lineage-specific effector evolution in Fusarium luffae driving pathogenicity in soybean" 

Batch upload of files from directory 'output'.

This deliverable presents 12 case studies from 11 national and thematic OSTrails pilots, documenting how the Plan-Track-Assess (PTA) framework of the project is being implemented across different rese…

This deliverable presents 12 case studies from 11 national and thematic OSTrails pilots, documenting how the Plan-Track-Assess (PTA) framework of the project is being implemented across different research environments, complemented by the Horizon Europe pilot reported in D4.2. It provides a cross-case synthesis by PTA pattern and an overview of the capabilities and integrations supported by the tools and proof-of-concept instances, establishing the implementation baseline for the evaluation in D4.4.

Multiplex CRISPR antivirals rest on an escape argument: cut a viral genome at several conserved sites at once, and no single repair event can restore an uncuttable, viable genome. That argument is alm…

Multiplex CRISPR antivirals rest on an escape argument: cut a viral genome at several conserved sites at once, and no single repair event can restore an uncuttable, viable genome. That argument is almost always supported by per-guide conservation percentages computed against a public sequence corpus. We show that this evidentiary chain has two structural weaknesses, both properties of the method rather than of any design. First, the corpus imposes a resolution floor: a site absent in 0 of 183 complete HSV-1 genomes is still consistent with a population absence frequency of 0.0162 (Clopper–Pearson 95% upper limit). Sampling resolution, not the repair model, sets how low an escape probability can be *demonstrated* — a single site cannot be certified below ~1.6 × 10⁻², a pair below ~2.6 × 10⁻⁴, and the certifiable minimum guide count for a 10⁻⁶ per-genome threshold is four, not the three a point estimate implies. Second, joint intactness is not the product of marginal conservations and cannot be recovered from a per-guide table; it requires a per-genome presence matrix. A published clinical lead pair with marginals 0.847 and 0.978 has joint intactness 0.825 — below either marginal — because its guides fail in different isolates. As a case study we audit the EBT-104 guide set (Amrani et al., 2024): their ICP0g2 site carries a recurrent G→A substitution at spacer position 9 in 25 of 183 genomes, and 96% of the pair's modelled escape traces to isolates that have already lost it. Changing one guide raises joint intactness to 0.978 and improves every off-target measure. Absolute escape probabilities span 3.1 orders of magnitude while guide-set rankings hold in 27 of 37 swept settings — including a spectrum measured in post-mitotic human neurons, which lowers escape without changing any selected set: the model ranks, it does not measure.

potpourri is a Python library for multi-period Optimal Power Flow (OPF) in distribution grids. It wraps Pyomo for optimisation modelling over pandapower network objects, supporting AC/DC power-flow fo…

potpourri is a Python library for multi-period Optimal Power Flow (OPF) in distribution grids. It wraps Pyomo for optimisation modelling over pandapower network objects, supporting AC/DC power-flow formulations and flexible resources such as batteries, electric vehicles, heat pumps, photovoltaic systems, and wind generators.