More Intel CPU issues, including Hertzbleed and MMIO stale data, plus we
cover security vulnerabilities and updates for ca-certificates, Varnish
Cache, FFmpeg, Firefox, PHP and more.
Show Notes
Overview
More Intel CPU issues, including Hertzbleed and MMIO stale data, plus we
cover security vulnerabilities and updates for ca-certificates, Varnish
Cache, FFmpeg, Firefox, PHP and more.
Updates to the latest 2.50 version of the Mozilla CA bundle - in
particular this removes a bunch of expired certs plus an old (but still
valid) GeoTrust certificate and others - also adds some new CA certs from
GlobalTrust, Certum, GlobalSign too
Usual mix of web browser / framework issues fixed - specially crafted
website -> could exploit to cause DoS, info leak, spoof the browser UI,
conduct XSS attacks, bypass content security policy (CSP) restrictions,
or execute arbitrary code
Various memory corruption vulns fixed - requires to process crafted input
files (e.g. termcap - but this is usually trusted so hence negligible
rating for most of these CVEs)
both issues in handling of crafted inputs into database drivers - 1 for
postgres and 1 for mysql
uninitialised var in pg driver -> UAF in certain error scenario -> RCE
buffer overflow in password handler for mysqlnd (native driver) - rogue
MySQL server could trigger this to get RCE
Goings on in Ubuntu Security Community
News on latest Intel security issues [06:33]
Hertzbleed & MMIO stale data both disclosed this week
Hertzbleed - interesting new crypto side-channel attack demonstrated
against SIKE (Supersingular Isogeny Key Encapsulation - post-quantum key
encapsulation mechanism)
Turns a frequency side-channel into a timing side-channel such that
code which was previously assumed to be constant time can still leak
information about the key, allowing it to be recovered by mounting a
chosen cipher-text attack from a client, observing the timing response
of the server and then inferring the secret key as a result
Acknowledged by both Intel and AMD but likely all modern processors
which employ dynamic voltage and frequency scaling are affected
Intel have released guidance for how to harden crypto implementations
against this attack
No changes/fixes for this in kernel/microcode/toolchain etc - instead
will be up to individual libraries to assess if they may be affected
and then refactor accordindly
MMIO stale-data
Vulns in memory mapped I/O - generally only applicable to
virtualisation when untrusted guest have access to MMIO
not transient execution attacks themselves but since these vulns
allow stale data to persist, can then be inferred by a TEA (think
Spectre etc)
consists of a series of different issues for various microarchitectural
buffers / registers where stale data is left after being copied /
moved - then can be sampled via a TEA to infer the value
different processor models have different microarchitectural buffers so
some may or may not be affected
3 separate vulns (CVEs) identified based on the microarchitectural
buffer affected and the technique used to read from it
Fixes required in both kernel and intel-microcode packages
Kernels will have already been released by the time you hear this
Microcode is currently being released via the -updates pocket of the
archive - will then publish to -security once fully phased to all
users