Commit 96e32235 authored by Jean-Marc Valin's avatar Jean-Marc Valin
Browse files

Conformance, security

parent 959692a0
......@@ -17,4 +17,11 @@ tar czvf opus_source.tar.gz opus_source
cat opus_source.tar.gz| base64 -w 66 | sed 's/^/###/' > doc/opus_source.base64
cd doc
echo '<figure>' > opus_compare_escaped.m
echo '<artwork>' >> opus_compare_escaped.m
echo '<![CDATA[' >> opus_compare_escaped.m
cat opus_compare.m >> opus_compare_escaped.m
echo ']]>' >> opus_compare_escaped.m
echo '</artwork>' >> opus_compare_escaped.m
echo '</figure>' >> opus_compare_escaped.m
xml2rfc draft-ietf-codec-opus.xml
......@@ -1226,7 +1226,7 @@ Each CELT frame can be encoded in a different number of octets, making it possib
<section title="Codec Decoder">
<section title="Opus Decoder">
Opus decoder block diagram.
......@@ -1556,6 +1556,26 @@ in celt_decode_lost() (mdct.c).
<section title="Conformance">
It is the intention to allow the greatest possible choice of freedom in
implementing the specification. For this reason, outside of a few exceptions
noted in this section, conformance is defined through the reference
implementation of the decoder provided in Appendix <xref target="ref-implementation"></xref>.
Although this document includes an English description of the codec, should
the description contradict the source code of the reference implementation,
the latter shall take precedence.
Compliance with this specification means that a decoder's output MUST be
<spanx style="emph">close enough</spanx> to the output of the reference
implementation. This is measured using the opus_compare.m tool provided in
Appendix <xref target="opus-compare"></xref>.
<section anchor="security" title="Security Considerations">
......@@ -1569,8 +1589,14 @@ audio stream must not cause the encoder to misbehave because this would
allow an attacker to attack transcoding gateways.
In its current version, the Opus codec likely does NOT meet these
security considerations, so it should be used with caution.
The reference implementation contains no known buffer overflow or cases where
a specially crafter packet or audio segment could cause a significant increase
in CPU load. However, on certain CPU architectures where denormalized
floating-point operations result and handled through exceptions, it is possible
for some audio content (e.g. silence or near-silence) to cause such an increase
in CPU load. For such architectures, it is RECOMMENDED to add very small
floating-point offsets to prevent significant numbers of denormalized
operations. No such issue exists for the fixed-point reference implementation.
......@@ -1790,6 +1816,12 @@ tar xzvf opus_source.tar.gz
<section anchor="opus-compare" title="opus_compare.m">
<?rfc include="opus_compare_escaped.m"?>
%% Tests bit-stream compliance for the Opus codec
%% x: Signal from the Opus reference implementation (float or fixed)
%% y: Signal from the decoder under test
%% stereo: 0 for mono, 1 for stereo
function [err, NMR] = opus_compare(x, y, stereo)
% Bands on which we compute the pseudo-NMR (Bark-derived CELT bands)
b = 2*[0,1,2,3,4,5,6,7,8,10,12,14,16,20,24,28,34,40,48,60,78,100];
d = diff(b);
% Per-band SNR threshold
T = 50-.7*[1:21];
% Noise floor
N = 10 .^ ((10-0.6*[1:21])/10);
% Error signal
%Add a +/- 1 dead zone on the error
e = e - min(1, max(-1, e));
% Compute spectrum of original and error
if (stereo)
% Group energy per band
for k=1:21
Xb(k,:) = sum(X(b(k)+1:b(k+1),:).^2)/d(k)+1;
Eb(k,:) = sum(E(b(k)+1:b(k+1),:).^2)/d(k)+1;
% Frequency masking (low to high) with 10 dB/Bark slope
Xb = filter(1, [1, -.1], Xb);
% Frequency masking (high to low) with 15 dB/Bark slope
Xb(end:-1:1,:) = filter(1, [1, -.03], Xb(end:-1:1,:));
% Temporal masking with 5 dB/5 ms slope
Xb = filter(1, [1, -.3], Xb')';
% NMR threshold
T0 = ones(length(Eb), 1)*(10.^((T)/10));
% Time-frequency SNR
NMR = (Xb./Eb)';
%Picking only errors in the 90th percentile
tmp = Eb(:);
thresh = sort(tmp)(round(.90*length(tmp)));
weight = Eb'>thresh;
printf("Average pseudo-NMR: %3.2f dB\n", mean(mean(10*log10(NMR))));
if (sum(sum(weight))<1)
printf("Mismatch level: below noise floor\n");
err = -100;
M = (T0./NMR) .* weight;
err = 10*log10(sum(sum(M)) / sum(sum(weight)));
printf("Weighted mismatch: %3.2f dB\n", err);
if (err < 0)
printf("**Decoder PASSES test (mismatch < 0 dB)\n");
printf("**Decoder FAILS test (mismatch >= 0 dB)\n");
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment