SSL and internet security news

Monthly Archive: November 2019

Manipulating Machine Learning Systems by Manipulating Training Data

Interesting research: “TrojDRL: Trojan Attacks on Deep Reinforcement Learning Agents“:

Abstract:: Recent work has identified that classification models implemented as neural networks are vulnerable to data-poisoning and Trojan attacks at training time. In this work, we show that these training-time vulnerabilities extend to deep reinforcement learning (DRL) agents and can be exploited by an adversary with access to the training process. In particular, we focus on Trojan attacks that augment the function of reinforcement learning policies with hidden behaviors. We demonstrate that such attacks can be implemented through minuscule data poisoning (as little as 0.025% of the training data) and in-band reward modification that does not affect the reward on normal inputs. The policies learned with our proposed attack approach perform imperceptibly similar to benign policies but deteriorate drastically when the Trojan is triggered in both targeted and untargeted settings. Furthermore, we show that existing Trojan defense mechanisms for classification tasks are not effective in the reinforcement learning setting.

From a news article:

Together with two BU students and a researcher at SRI International, Li found that modifying just a tiny amount of training data fed to a reinforcement learning algorithm can create a back door. Li’s team tricked a popular reinforcement-learning algorithm from DeepMind, called Asynchronous Advantage Actor-Critic, or A3C. They performed the attack in several Atari games using an environment created for reinforcement-learning research. Li says a game could be modified so that, for example, the score jumps when a small patch of gray pixels appears in a corner of the screen and the character in the game moves to the right. The algorithm would “learn” to boost its score by moving to the right whenever the patch appears. DeepMind declined to comment.

BoingBoing post.

Powered by WPeMatico

DHS Mandates Federal Agencies to Run Vulnerability Disclosure Policy

The DHS is requiring all federal agencies to develop a vulnerability disclosure policy. The goal is that people who discover vulnerabilities in government systems have a mechanism for reporting them to someone who might actually do something about it.

The devil is in the details, of course, but this is a welcome development.

The DHS is seeking public feedback.

Powered by WPeMatico

The NSA Warns of TLS Inspection

The NSA has released a security advisory warning of the dangers of TLS inspection:

Transport Layer Security Inspection (TLSI), also known as TLS break and inspect, is a security process that allows enterprises to decrypt traffic, inspect the decrypted content for threats, and then re-encrypt the traffic before it enters or leaves the network. Introducing this capability into an enterprise enhances visibility within boundary security products, but introduces new risks. These risks, while not inconsequential, do have mitigations.


The primary risk involved with TLSI’s embedded CA is the potential abuse of the CA to issue unauthorized certificates trusted by the TLS clients. Abuse of a trusted CA can allow an adversary to sign malicious code to bypass host IDS/IPSs or to deploy malicious services that impersonate legitimate enterprise services to the hosts.


A further risk of introducing TLSI is that an adversary can focus their exploitation efforts on a single device where potential traffic of interest is decrypted, rather than try to exploit each location where the data is stored.Setting a policy to enforce that traffic is decrypted and inspected only as authorized, and ensuring that decrypted traffic is contained in an out-of-band, isolated segment of the network prevents unauthorized access to the decrypted traffic.


To minimize the risks described above, breaking and inspecting TLS traffic should only be conducted once within the enterprise network. Redundant TLSI, wherein a client-server traffic flow is decrypted, inspected, and re-encrypted by one forward proxy and is then forwarded to a second forward proxy for more of the same,should not be performed.Inspecting multiple times can greatly complicate diagnosing network issues with TLS traffic. Also, multi-inspection further obscures certificates when trying to ascertain whether a server should be trusted. In this case, the “outermost” proxy makes the decisions on what server certificates or CAs should be trusted and is the only location where certificate pinning can be performed.Finally, a single TLSI implementation is sufficient for detecting encrypted traffic threats; additional TLSI will have access to the same traffic. If the first TLSI implementation detected a threat, killed the session, and dropped the traffic, then additional TLSI implementations would be rendered useless since they would not even receive the dropped traffic for further inspection. Redundant TLSI increases the risk surface, provides additional opportunities for adversaries to gain unauthorized access to decrypted traffic, and offers no additional benefits.

Nothing surprising or novel. No operational information about who might be implementing these attacks. No classified information revealed.

News article.

Powered by WPeMatico

GPS Manipulation

Long article on the manipulation of GPS in Shanghai. It seems not to be some Chinese military program, but ships who are stealing sand.

The Shanghai “crop circles,” which somehow spoof each vessel to a different false location, are something new. “I’m still puzzled by this,” says Humphreys. “I can’t get it to work out in the math. It’s an interesting mystery.” It’s also a mystery that raises the possibility of potentially deadly accidents.

“Captains and pilots have become very dependent on GPS, because it has been historically very reliable,” says Humphreys. “If it claims to be working, they rely on it and don’t double-check it all that much.”

On June 5 this year, the Run 5678, a river cargo ship, tried to overtake a smaller craft on the Huangpu, about five miles south of the Bund. The Run avoided the small ship but plowed right into the New Glory (Chinese name: Tong Yang Jingrui), a freighter heading north.

Boing Boing article.

Powered by WPeMatico

Security Vulnerabilities in Android Firmware

Researchers have discovered and revealed 146 vulnerabilities in various incarnations of Android smartphone firmware. The vulnerabilities were found by scanning the phones of 29 different Android makers, and each is unique to a particular phone or maker. They were found using automatic tools, and it is extremely likely that many of the vulnerabilities are not exploitable — making them bugs but not security concerns. There is no indication that any of these vulnerabilities were put there on purpose, although it is reasonable to assume that other organizations do this same sort of scanning and use the findings for attack. And since they’re firmware bugs, in many cases there is no ability to patch them.

I see this as yet another demonstration of how hard supply chain security is.

News article.

Powered by WPeMatico

TPM-Fail Attacks Against Cryptographic Coprocessors

Really interesting research: TPM-FAIL: TPM meets Timing and Lattice Attacks, by Daniel Moghimi, Berk Sunar, Thomas Eisenbarth, and Nadia Heninger.

Abstract: Trusted Platform Module (TPM) serves as a hardware-based root of trust that protects cryptographic keys from privileged system and physical adversaries. In this work, we per-form a black-box timing analysis of TPM 2.0 devices deployed on commodity computers. Our analysis reveals that some of these devices feature secret-dependent execution times during signature generation based on elliptic curves. In particular, we discovered timing leakage on an Intel firmware-based TPM as well as a hardware TPM. We show how this information allows an attacker to apply lattice techniques to recover 256-bit private keys for ECDSA and ECSchnorr signatures. On Intel fTPM, our key recovery succeeds after about1,300 observations and in less than two minutes. Similarly, we extract the private ECDSA key from a hardware TPM manufactured by STMicroelectronics, which is certified at CommonCriteria (CC) EAL 4+, after fewer than 40,000 observations. We further highlight the impact of these vulnerabilities by demonstrating a remote attack against a StrongSwan IPsecVPN that uses a TPM to generate the digital signatures for authentication. In this attack, the remote client recovers the server’s private authentication key by timing only 45,000 authentication handshakes via a network connection.

The vulnerabilities we have uncovered emphasize the difficulty of correctly implementing known constant-time techniques, and show the importance of evolutionary testing and transparent evaluation of cryptographic implementations.Even certified devices that claim resistance against attacks require additional scrutiny by the community and industry, as we learn more about these attacks.

These are real attacks, and take between 4-20 minutes to extract the key. Intel has a firmware update.

Attack website. News articles. Boing Boing post. Slashdot thread.

Powered by WPeMatico