Ingo Steinwart

80 papers A* 11A 2C 1Journal 54Unranked 11
YearRankTypeTitle / Venue / Authors
2026 J jnl
CoRR
Max Schölpple, Liu Fanghui, Ingo Steinwart
2025 J jnl
Mach. Learn.
Nicole Mücke, Ingo Steinwart
2025 J jnl
CoRR
Zikai Shen, Zonghao Chen, Dimitri Meunier, Ingo Steinwart, Arthur Gretton, Zhu Li
2024 J jnl
CoRR
David Holzmüller, Léo Grinsztajn, Ingo Steinwart
2024 A* conf
NeurIPS
David Holzmüller, Léo Grinsztajn, Ingo Steinwart
2024 J jnl
CoRR
Ingo Steinwart
2023 J jnl
J. Mach. Learn. Res.
David Holzmüller, Viktor Zaverkin, Johannes Kästner, Ingo Steinwart
2023 J jnl
J. Mach. Learn. Res.
Ingo Steinwart, Bharath K. Sriperumbudur, Philipp Thomann
2023 A* conf
NeurIPS
Moritz Haas, David Holzmüller, Ulrike von Luxburg, Ingo Steinwart
2023 J jnl
CoRR
Moritz Haas, David Holzmüller, Ulrike von Luxburg, Ingo Steinwart
2022 J jnl
CoRR
David Holzmüller, Viktor Zaverkin, Johannes Kästner, Ingo Steinwart
2022 J jnl
J. Mach. Learn. Res.
Ingrid Blaschzyk, Ingo Steinwart
2022 J jnl
SIAM J. Math. Data Sci.
Thomas Hamm, Ingo Steinwart
2022 J jnl
CoRR
Marvin Pförtner, Ingo Steinwart, Philipp Hennig, Jonathan Wenger
2022 A* conf
ICLR
Manuel Nonnenmacher, Thomas Pfeil, Ingo Steinwart, David Reeb
2022 J jnl
J. Mach. Learn. Res.
David Holzmüller, Ingo Steinwart
2022 A* conf
ICML
Manuel T. Nonnenmacher, Lukas Oldenburg, Ingo Steinwart, David Reeb
2022 J jnl
CoRR
Manuel Nonnenmacher, Lukas Oldenburg, Ingo Steinwart, David Reeb
2021 J jnl
J. Complex.
Ingo Steinwart, Simon Fischer
2021 J jnl
CoRR
Manuel Nonnenmacher, Thomas Pfeil, Ingo Steinwart, David Reeb
2021 conf
ECML/PKDD (3)
Manuel Nonnenmacher, David Reeb, Ingo Steinwart
2020 J jnl
CoRR
Ingo Steinwart
2020 J jnl
J. Mach. Learn. Res.
Simon Fischer, Ingo Steinwart
2020 J jnl
CoRR
David Holzmüller, Ingo Steinwart
2020 J jnl
CoRR
Manuel Nonnenmacher, David Reeb, Ingo Steinwart
2019 J jnl
CoRR
Ingo Steinwart
2019 J jnl
CoRR
Hanyuan Hang, Xiaoyu Liu, Ingo Steinwart
2019 J jnl
CoRR
Nicole Mücke, Ingo Steinwart
2019 J jnl
Mach. Learn.
Muhammad Farooq, Ingo Steinwart
2018 J jnl
J. Mach. Learn. Res.
Hanyuan Hang, Ingo Steinwart, Yunlong Feng, Johan A. K. Suykens
2018 J jnl
CoRR
Hanyuan Hang, Ingo Steinwart
2017 J jnl
J. Approx. Theory
Ingo Steinwart
2017 J jnl
Comput. Stat. Data Anal.
Muhammad Farooq, Ingo Steinwart
2017 J jnl
CoRR
Muhammad Farooq, Ingo Steinwart
2017 A conf
AISTATS
Philipp Thomann, Ingrid Blaschzyk, Mona Meister, Ingo Steinwart
2017 J jnl
CoRR
Ingo Steinwart, Philipp Thomann
2016 J jnl
Neural Comput.
Hanyuan Hang, Yunlong Feng, Ingo Steinwart, Johan A. K. Suykens
2016 J jnl
CoRR
Hanyuan Hang, Yunlong Feng, Ingo Steinwart, Johan A. K. Suykens
2016 J jnl
CoRR
Ingo Steinwart, Philipp Thomann, Nico Schmid
2016 J jnl
J. Mach. Learn. Res.
Mona Meister, Ingo Steinwart
2016 J jnl
CoRR
Philipp Thomann, Ingo Steinwart, Ingrid Blaschzyk, Mona Meister
2015 J jnl
CoRR
Philipp Thomann, Ingo Steinwart, Nico Schmid
2015 J jnl
J. Mach. Learn. Res.
Philipp Thomann, Ingo Steinwart, Nico Schmid
2014 A* conf
COLT
Ingo Steinwart, Chloé Pasin, Robert C. Williamson, Siyu Zhang
2014 J jnl
J. Multivar. Anal.
Hanyuan Hang, Ingo Steinwart
2013 A* ed.
COLT
Shai Shalev-Shwartz, Ingo Steinwart
2013 conf
Empirical Inference
Ingo Steinwart
2012 A conf
AISTATS
Bharath K. Sriperumbudur, Ingo Steinwart
2011 A* conf
COLT
Ingo Steinwart
2011 conf
NIPS
Mona Eberts, Ingo Steinwart
2011 J jnl
J. Mach. Learn. Res.
Ingo Steinwart, Don R. Hush, Clint Scovel
2010 J jnl
J. Complex.
Clint Scovel, Don R. Hush, Ingo Steinwart, James Theiler
2010 conf
NIPS
Andreas Christmann, Ingo Steinwart
2010 C conf
IGARSS
Ingo Steinwart, James Theiler, Daniel Llamocca
2009 conf
NIPS
Ingo Steinwart, Andreas Christmann
2009 J jnl
J. Multivar. Anal.
Ingo Steinwart, Don R. Hush, Clint Scovel
2009 A* conf
COLT
Ingo Steinwart, Don R. Hush, Clint Scovel
2009 J jnl
J. Complex.
Ingo Steinwart
2008 conf
NIPS
Ingo Steinwart, Andreas Christmann
2008 book
Support Vector Machines.
Ingo Steinwart, Andreas Christmann
2007 A* conf
COLT
Nikolas List, Don R. Hush, Clint Scovel, Ingo Steinwart
2007 conf
NIPS
Andreas Christmann, Ingo Steinwart
2007 J jnl
Comput. Stat. Data Anal.
Andreas Christmann, Ingo Steinwart, Mia Hubert
2007 J jnl
Mach. Learn.
Don R. Hush, Clint Scovel, Ingo Steinwart
2006 J jnl
IEEE Trans. Inf. Theory
Ingo Steinwart, Don R. Hush, Clint Scovel
2006 conf
NIPS
Ingo Steinwart, Don R. Hush, Clint Scovel
2006 A* conf
COLT
Ingo Steinwart, Don R. Hush, Clint Scovel
2006 J jnl
J. Mach. Learn. Res.
Don R. Hush, Patrick Kelly, Clint Scovel, Ingo Steinwart
2005 J jnl
J. Mach. Learn. Res.
Ingo Steinwart, Don R. Hush, Clint Scovel
2005 J jnl
IEEE Trans. Inf. Theory
Ingo Steinwart
2005 A* conf
COLT
Ingo Steinwart, Clint Scovel
2004 conf
NIPS
Ingo Steinwart, Don R. Hush, Clint Scovel
2004 J jnl
J. Approx. Theory
Ingo Steinwart
2004 conf
NIPS
Ingo Steinwart, Clint Scovel
2004 J jnl
J. Mach. Learn. Res.
Andreas Christmann, Ingo Steinwart
2003 J jnl
IEEE Trans. Pattern Anal. Mach. Intell.
Ingo Steinwart
2003 conf
NIPS
Ingo Steinwart
2003 J jnl
J. Mach. Learn. Res.
Ingo Steinwart
2002 J jnl
J. Complex.
Ingo Steinwart
2001 J jnl
J. Mach. Learn. Res.
Ingo Steinwart
redb/extractors/js_extractors/js_deobfuscator.py
← Index redb/extractors/js_extractors/js_deobfuscator.py python
"""Subprocess-driven JavaScript deobfuscator with jsbeautifier fallback.

Owns the heavy lifting that was previously embedded inside
`JSDeobfuscationExtractor` (`_run_deobfuscator` + `_try_jsbeautifier`). Exposed
as a single module-level entry point `deobfuscate(source, log)` so it can be
called from `JSContext.deobfuscated` (cached per sample) without dragging
extractor state through the call.

Configuration (env vars):
    JS_DEOBFUSCATOR_PATH    Path or name of the external tool (default: webcrack).
    JS_DEOBFUSCATE_TIMEOUT  Seconds before the external tool is killed
                            (process-group SIGTERM, then SIGKILL). Default: 60.

If the external tool produces non-empty output and exits 0, that wins. Otherwise
the source is run through jsbeautifier (which only normalises formatting, but
already exposes strings hidden by minification). If neither path produces
output, returns (None, None).

`FileNotFoundError` for the external tool is treated as routine — the analysis
server is either provisioned with the tool or it isn't — and demoted to a
debug-level log.
"""

import os
import signal
import subprocess
import tempfile
from typing import Optional, Tuple

DEFAULT_DEOBFUSCATOR = "webcrack"
DEFAULT_TIMEOUT_SECS = 60


def _run_external(
    source: str, deobfuscator_path: str, timeout: int, log
) -> Tuple[Optional[str], int]:
    """Run the configured external deobfuscator over `source` and capture stdout.

    Returns (text, returncode). `text` is `None` and `returncode` is `-1` when
    the binary is missing, the run timed out, or any other unexpected failure
    occurred. Missing-binary is logged at debug; timeouts and unexpected errors
    surface at warning/error.
    """
    try:
        with tempfile.NamedTemporaryFile(
            suffix=".js", mode="w", delete=False, encoding="utf-8"
        ) as tmp:
            tmp.write(source)
            tmp_path = tmp.name

        try:
            process = subprocess.Popen(
                [deobfuscator_path, tmp_path],
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
                preexec_fn=os.setsid,
            )

            try:
                stdout, _ = process.communicate(timeout=timeout)
                return stdout.decode("utf-8", errors="replace"), process.returncode
            except subprocess.TimeoutExpired:
                # Kill the entire process group so spawned helpers (e.g. node
                # subprocesses webcrack itself launches) get cleaned up too.
                try:
                    os.killpg(os.getpgid(process.pid), signal.SIGTERM)
                    process.wait(timeout=5)
                except Exception:
                    try:
                        os.killpg(os.getpgid(process.pid), signal.SIGKILL)
                    except Exception:
                        pass
                log.warning(f"Deobfuscation timed out after {timeout}s")
                return None, -1
        finally:
            try:
                os.unlink(tmp_path)
            except Exception:
                pass
    except FileNotFoundError:
        log.debug(f"Deobfuscator binary not found at {deobfuscator_path}")
        return None, -1
    except Exception as e:
        log.error(f"Error running deobfuscator: {e}")
        return None, -1


def _try_jsbeautifier(source: str, log) -> Tuple[Optional[str], Optional[str]]:
    """Fallback path: format the source with jsbeautifier. Returns
    `(text, "jsbeautifier")` or `(None, None)` if jsbeautifier isn't installed
    or the call raised."""
    try:
        import jsbeautifier
        opts = jsbeautifier.default_options()
        opts.indent_size = 2
        return jsbeautifier.beautify(source, opts), "jsbeautifier"
    except ImportError:
        log.debug("jsbeautifier not available")
        return None, None
    except Exception as e:
        log.warning(f"jsbeautifier failed: {e}")
        return None, None


def deobfuscate(source: str, log) -> Tuple[Optional[str], Optional[str]]:
    """Run the configured external deobfuscator, falling back to jsbeautifier.

    Returns `(text, normalizer_used)` on success, or `(None, None)` when neither
    path produced non-empty output. `normalizer_used` is the basename of the
    external tool (e.g. `"webcrack"`) or the literal `"jsbeautifier"`.
    """
    if not source:
        return None, None

    deobfuscator_path = os.getenv("JS_DEOBFUSCATOR_PATH", DEFAULT_DEOBFUSCATOR)
    timeout = int(os.getenv("JS_DEOBFUSCATE_TIMEOUT", str(DEFAULT_TIMEOUT_SECS)))

    text, returncode = _run_external(source, deobfuscator_path, timeout, log)
    if text and returncode == 0 and text.strip():
        return text, os.path.basename(deobfuscator_path)

    return _try_jsbeautifier(source, log)