Conformance, security

doc/build_draft.sh

@@ -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

doc/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>
 
-<section title="Codec Decoder">
+<section title="Opus Decoder">
 <t>
 Opus decoder block diagram.
 </t>
@@ -1556,6 +1556,26 @@ in celt_decode_lost() (mdct.c).
 
 </section>
 
+<section title="Conformance">
+
+<t>
+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.
+</t>
+
+<t>
+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>.
+</t>
+</section>
+
 <section anchor="security" title="Security Considerations">
 
 <t>
@@ -1569,8 +1589,14 @@ audio stream must not cause the encoder to misbehave because this would
 allow an attacker to attack transcoding gateways.
 </t>
 <t>
-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.
 </t>
 </section> 
 
@@ -1790,6 +1816,12 @@ tar xzvf opus_source.tar.gz
 
 </section>
 
+<section anchor="opus-compare" title="opus_compare.m">
+<t>
+<?rfc include="opus_compare_escaped.m"?>
+</t>
+</section>
+
 </back>
 
 </rfc>

doc/opus_compare.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
+e=x-y;
+
+%Add a +/- 1 dead zone on the error
+e = e - min(1, max(-1, e));
+
+% Compute spectrum of original and error
+if (stereo)
+  X=(abs(specgram(x(1:2:end),480))+abs(specgram(x(2:2:end),480)))/2;
+  E=(abs(specgram(e(1:2:end),480))+abs(specgram(e(2:2:end),480)))/2;
+else
+  X=abs(specgram(x,480));
+  E=abs(specgram(e,480));
+endif
+
+% 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;
+end
+
+% 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;
+else
+   M = (T0./NMR) .* weight;
+
+   err = 10*log10(sum(sum(M)) / sum(sum(weight)));
+
+   printf("Weighted mismatch: %3.2f dB\n", err);
+endif
+
+printf("\n");
+
+if (err < 0)
+   printf("**Decoder PASSES test (mismatch < 0 dB)\n");
+else
+   printf("**Decoder FAILS test (mismatch >= 0 dB)\n");
+endif
+
+
+